hr_gamification should be adding hr groups to the security of gamification.
Creating the group base.group_hr_manager in gamification in case hr is not
installed is confusing and redundant with existing manager groups.

Creating a external id in a different module than the one where the record is
declared is a bad practice.
Setting a correct module is both important for a correct uninstallation of the
module (which uses the module to decide which records to remove) and for the
translations (records tagged with 'base' module needs to be declared in the
base.pot file).

one2many, many2many, text and html fields should take all space in a form view
for a better rendering instead of taking only 2 col or not having the label in
front the field.

When a model has no form defined, a default view is generated displaying all
fields of the model.
The one2many and many2many fields were not displayed on the form, probably for
aesthetic or historical (2008) reasons.

The snippet section can have a color or a background. Previous commit
added the functionnality for its innercolumn too... but having a bg
image on top of another one does not make sense, so only keep the
color option for columns.

- s_references
    * Replace demo images (lighter ones)
    * Move title to left
    * Review responsiveness
- s_big_message
    * Add background
    * Use grid system

* Replace demo blog post covers
* Add xml ID to demo images so they can be easily replaced by themes
* Remove unused images

=======
Purpose
=======

The company form should be heavily simplified. It's complex to have some settings on the company form, and others on the Settings menu of the related app. It would be much easier to have all settings in Settings menu (res.config) and nothing on the company, even if some of the Settings are multi company. (Stored on the company but set from the Settings menu).

Historically, developers had to use the `eval` attribute to set
a False value for boolean fields in XML data files:

    <field name="active" eval="False"/>

This is a source of errors for beginners, who tend to provide
a string value directly, as for char/text fields.
Unfortunately string values always evaluated to `True` booleans:

    <field name="active">False</field> <!-- active == true! -->

This patch adds detection of some well-known falsy strings
(case-insensitive): `0`, `false`, `off`, as similarly supported
for the `noupdate` and `forcecreate` attributes of XML data records.

Closes #13152

* The debug manager contains a feature to launch tours, adapt it
to new web_tour.tour system
* The website help menu allowed to relaunch tutorial tour in debug
mode. This feature is not adapted to new tour system and was useless
in the first place (debug mode in frontend is barely used anyway).
-> remove the website help menu and the feature

(unlaunch, need to create an new test for menu)

Since res.users is an inherits of res.partner, the onchange that
should be triggered by tz to update tz_offset is not triggered.

Considering tz_offset is never used on res_partner and is only
used by the timezone_mistmatch widget which compares the offset
with the one of the browser timezone, which only makes sense if
you are a user, we could have completely moved this field on
res.users. That being said, it feels strange to have a tz field
defined on partner but having to be a user to compute the actual
timezon offset, which doesn't need any additional information.

For this reason, we chose to duplicate it until either we decide
to truly move it in the next version, or the orm is fixed to
correctly trigger the onchanges of the inherits models.

This patch could be applied in 9.0 if needed, but considering
nobody ever complained about this issue and the "workaround"
nature of this fix, we decided to apply it in master only.

and the bad computed & required & default combinaison

stage_id is comptued and based on stage_ids
The default value should be put on stage_ids, not stage_id, it has the same
effect but avoids an unecessary chain update.

Which is useless when there is no inverse method

The account.move has already self.date for the date field (cf the method
`_prepare_reconciliation_move`). As the field date on the account.move.line is
a related, it makes useless updates to write it here.

Same issue for journal_id field

ref is a related set on the account.move. Writting on the line set the same
reference for all the lines and the parent move (the one of the last line).
Instead put it on the move if there is only one reference or do not put a wrong
one if we can't determine one.

Which does not make sense. it's the computation that should set the value.
Also the default is no longer needed.

journal_id is a related field that was set just before

Avoid setting a value for a related field as it will trigger onchange update.
This code was relevant when the date was different on the move and the line but
is no longer needed.

Remove all the `if vals.get(mandatory_field)` to simplify this method and remove
dead code or condition always true.

The field journal_id and date are both required on account.move and are stored
related on account.move.line.
Setting the required on the account.move.line allows to make it required in the
views or for the ORM but in practice it will be same as it's required on the
parent.

Making it required is a performance killer as it forces to specify a value for
the create which will retrigger an update on the parent move.
This means that adding a line with a date value on an account.move with already
10 lines will trigger 11 updates (date of the new line -> date of the move ->
date on the 10 existing lines), even if the value has not changed.

Currently some customers receive the automatic notification 'Quotation
viewed by customer'. This should not be the case as this notifciation
is purely internal and should only be sent to the quote responsible.

Notification is now a note with the responsible being directly in the
recipients.

Commit 93394908 added the odoo variable
definition in the iframe template so the condition which checked if
odoo.define is available had to be adapted.

The 'purchase_ok' field has been moved into the 'purchase' module in
rev https://github.com/odoo/odoo/commit/e945b86d5a61e53a9104ab6cc4ed4a1c060587e4 and removed from 'product'
module in rev https://github.com/odoo/odoo/commit/fee03009cfcc41d108a79a2380912f21bc825d77 to fix a duplicated field.
But it should actually stay only in 'product' module because the field
is needed in the 'account' module which doesn't depend of 'purchase'.

So to fix it we set back the 'purchase_ok' field in the product module
as it was originally in the rev https://github.com/odoo/odoo/commit/e23e018285b17450af97cf93165704926f8248db

The nocontent message was never there when the kanban had no content
in grouped mode and the create button was muted even if no column can
be added. Also it was wrongly displayed in ungrouped mode.

The pre-processing of kanban qweb templates dropped the t-att(f)-class
of button and tags with a valid type attr (object, action, ...).

Workflows are removed and replaced by python code directly embeded in model
files. There are several reasons to do this

 * most of workflows are simple state machines. They simple go from state to
   state under simple conditions. Short circuits exist allowing to bypass
   part of the flow, like cancelling a record. However it is easier to
   understand a flow composed of several simple methods than having to
   understand the workflow in a separate xml file
 * workflows are hard to debug and modify. Each modification require a small
   migration. Moreover ensuring records are not blocked or that there does
   not exist unreachable states is hard to do.
 * having all business code in models helps understanding the business code.
 * most of our flows are quite simple and does not require a complex machinery
   to reflect it

Workflow engine itself is kept. Lot of people use it and it can be very
useful in some cases, notably in modeling more complex flows. Removing it
is therefore not planned in this task.

Commit 744cd64b re-added the removed
col system in the title snippet but did not respect the odoo structure
which is section > .container [> .row > .col-*] and put the container
class directly on the section. With this method, when changing the
background of the snippet, only the container part is filled
with background.

* On firefox, you had to click twice on a link to be able to start
typing text
* When selecting text which contained a <br/>, you were not able to
replacing it correctly (you had to click again before starting typing)
* When replacing a whole paragraph content which was below a title,
the text was not replaced and you had to click again and the title
transformed into a paragraph.
* ...

remove workflow and replace by python code.
note: hr.payslip states are messy, 4 states are defined, 3 are used
      and 5 exist.