After commit 45dae71b fixed the boot on Renesas boards with
linux >= 5.7 due to the change of the dtb name, the tests exercising
the NFS boot of the current release as the first stage broke since
they do not use `boot_commands` and thus didn't get the override.

To fix that, rather than constructing the path to the DTB from the
LAVA device name, which no longer match the board type with
linux >= 5.7, explicitly pass the DTB path for the first stage and the
release under test separately.

Signed-off-by: Emanuele Aina <emanuele.aina@collabora.com>

Signed-off-by: Emanuele Aina <emanuele.aina@collabora.com>

It turns out that LAVA really wants strings for the metadata entries and
didn't like nesting a dictionary there.

For this reason it was failing to submit jobs with this error:

    Submitting job group-apparmor-tpl.yaml: <Fault 400: "Problem with submitted job data: expected str for dictionary value @ data['metadata']['source']">

Unfortunately our custom `lava-submit.py` was silently eating the error
so the pipeline was marked as successful.

Signed-off-by: Emanuele Aina <emanuele.aina@collabora.com>

The `source:wip:` metadata can be used to indicate that a submission
came from a work-in-progress branch and that it should not be accounted
when checking the overall health of the projects.

For instance, the Phabricator bridge can use this information to avoid
opening bug tasks for failures coming from a WIP branch.

Signed-off-by: Emanuele Aina <emanuele.aina@collabora.com>

Introduce the variables below:

* source_ref
* source_commit
* source_pipeline
* source_job

Signed-off-by: Emanuele Aina <emanuele.aina@collabora.com>

In preparation for adding more metadata, make the `source` entry an
object rather than a single string.

Signed-off-by: Emanuele Aina <emanuele.aina@collabora.com>

Default to `null` rather than hardconding a possibly misleading source.

Signed-off-by: Emanuele Aina <emanuele.aina@collabora.com>

Keeping the LAVA job templates in `apertis-tests` has many problems.

Since `apertis-tests` is also a package, updates for stable releases
should go through `-updates` branches, but that plays badly with the
image building recipes as they only pull from the the main branch.

There's also an annoying disconnect about what the testcases here say
about where they should be run, and the actual groups used by the
templates.

By moving the LAVA templates here we can later improve them by
generating the groups based on the actual data in the testcases.

Signed-off-by: Emanuele Aina <emanuele.aina@collabora.com>