This merge of "rafael" branch includes several commits for:

  - Fixes for Functional Tests
  - Use Remote URL for Tutorial Related tests (see testFunctionalConfigurator)
  - Fixes for Consulting Configurator Workflow tests
  - Several tests for SVG Conversion and better SVG support

SVG has a special treatment on this case, because is the only format which
is a real image. The others formats are non-image cases.

When it is SVG, rely only on Image API Conversion.

Merged master into rafael branch

Conflicts:
	bt5/erp5_ui_test/bt/revision
	product/ERP5/bootstrap/erp5_core/bt/revision

Person, Organisation, Link person and Organisation and few others tests
were merged with the tutorials, due this some TESTONLY were archived.

This is surely the source of random duplicate uid problems. Indeed,
some test like testERP5Catalog using default extra_sql_connection_string_list
provided by slapos were using same connection string as other running tests

If a TextDocument is used as image (ie.: like in Web Illustration to SVG),
the conversion API should be from Image to prevent code duplications.

Transform the URL into Data URI in order to avoid ImageMagick and
rsvg-convert limitations. This API should be dropped as soon those external
tools can handle remote URLs (ie.: http, https ...) into SVG definitions.

It was included defensive code and a lot of tests to prevent unexpected
outcomes, like raise ConversionError if the download failures and do
not parse empty content.

This way, we can effectively kill all subprocesses. This worked for
slapproxy because Flask calls setsid internally, but slapgrid-* doesn't, so
they wouldn't be killed by killpg.

Fixes upgrade path from older erp5testnode version.

Note that it is useless to test simulation quantity in isSimulated
because even if it exists, we already failed to find a related SM.

This will be required to fix a possible race condition in causality_workflow

Otherwise, first test may fail if '--random_activity_priority=' option is used.

This fix is already applied to erp5.util.testnode, we duplicate
the fix. The code will move to library soon

The stop_condition() check is run once every beginning of loop tic, but if a loop
achieved a state of total failure, RuntimeError was raised before stop_condition()
had a chance to run.

So we call it again just one more time to see if we can avoid failing.

the code of some id generator can raise ValueError when objects
are not fully installed, and this is typically a case when the
compatibility code needs to run (like uid generation in the middle
of the installation of id generators objects)

This is used to get ongoing status even though there is no failures,
it allows to know what is going on on testnode side

This fix errors that could depend on environment (like python
version, mariadb version, etc)