As we now have a new debug mode 'tests' which load a new asset bundle
containing tour-test files, we moved those files to a new folder hierarchy.
That will clean the .js files trees.
Also, those files should be included in the new asset.

Basically, the .js tour files (not test) should be inside /static/src/js/tours
while .js tour test files (test=true) should be inside /static/tests/tours next
to QUnit tests, inside a tours folder.

+ test_new_api: don't run the test in debug assets

task-1934445
Comes with https://github.com/odoo/enterprise/pull/4281
Closes #33213

* portal, sale, website, website_blog, website_crm, website_event,
  website_event_sale, website_forum, website_hr_recruitment,
  website_sale, website_sale_wishlist

At last, that ugly JS module can be removed. Before this current PR, it
was still used to wait for "page ready", to initialize widgets on page
loading. Now, all can be done thanks to public root and public widgets.

web_editor.base was also still used in tours, which should not be
necessary anymore for the same reason, especially since the tour manager
waits for the public root naturally.

Part of https://github.com/odoo/odoo/pull/29442
task-1932066

Scss tour was broken due to recent task that shows only 2 new custom scss
files if not in edit mode. As the tests was trying to read `website.ui.scss`
that is now in debug mode in a filter, it would not be found.

Html tour was broken due to new editor that changed the save process.
For some reason, after the drag & drop of a snippet, clicking fast on save
would not fire RPCs as it would not be considered as dirty.
(Editor would then close, you would still see the modifications even if nothing
was saved on DB)

closes odoo/odoo#31067

Before this commit:
1. If we were editing a template with no xml_id, typically the case when
   editing a specific view (COW'd), the 'Template ID' label would be left empty
   as the view does not have an xml_id but only a key.
2. If we were editing a generic view, the template would be COW'd and replaced
   by the specific one created during the save.
   After the save, the page would be reload and will try to reopen the edited
   template (stored in url `res=123`). But as this template would be replaced
   by the COW'd view, the JS would crash trying to access an inexisting view.
   The HTML editor would then open in a very thin modal, barely editable.

Now:
1. Display the template's key instead of its xml_id. The key is basically a
   duplicate of the xml_id, mainly used in website.
2. If we are saving a generic view, we search for its newly created specific
   view to upload the URL hash before reloading.

task-1934279
Fix #30447

As this behavior was recently broken (fixed in #30403), this commit
adds a test tour for that flow.

task-1934279
Fix #30447

With the following view hierarchy:
   A
   |
   B' (oe_structure_ID view)

Before this fix, when saving multiple templates at once from the same hierarchy
the COW views (website) would not be saved correctly as the first call to
`write` would COW the parent view, then the second concurent call to `write`
would create the child specific view under the generic parent view.
Instead, the generic parent should have been cloned to a specific parent view
and then the child view should be COW under the specific parent.

  A    A'
  |    |
  B'   B*' (This one is the old B', without B' changes in the html editor)

Instead of

  A   A'
      |
      B'

Calling the `write` one after another will do the trick. Indeed, when the
second call is fired, the first call has already triggered the COW behavior.
Thus, the second call correctly retrieves the COW view.

Another issue might still appear, if the parent view write is sent first.
Indeed, the parent view will be COW, thus the specific child view (second write
waiting to be fired) will be deleted to be added on the new specific hierarchy.
Thus, that write will then raise an error since the view does not exist anymore.

task-1934279
Fix #30447