Add home-assistant to primer and remove old external primer

[jacobtylerwalls]

Type of Changes

	Type
✓	🔨 Refactoring

Description

Restore home-assistant to the new primer, and finally remove the old (external) primer. The old (check for fatal only) primer still runs on the stdlib.

We were waiting for the 3.7 primer job to take less time before doing this, but we fixed that in #8608, and we're dropping 3.7 soon.

The commit string was too large for the cache action (512 chars ceiling), so I shortened the SHAs it uses.

Removed keras, as it wasn't linting successfully in this context, unfortunately.

Successful run on my fork

Closes #5359

[codecov]

Codecov Report

Merging #8612 (a0db1b9) into main (a83137d) will not change coverage.
The diff coverage is n/a.

Additional details and impacted files

@@           Coverage Diff           @@
##             main    #8612   +/-   ##
=======================================
  Coverage   95.91%   95.91%           
=======================================
  Files         174      174           
  Lines       18416    18416           
=======================================
  Hits        17664    17664           
  Misses        752      752           

Impacted Files	Coverage Δ
pylint/testutils/_primer/primer_prepare_command.py	23.33% <ø> (ø)

[Pierre-Sassoulas]

Isn't keras failing in the "new" primer a valid reason to have it in the primer ? (If we only add projects that work from the get go we're selecting projects that are somewhat more compatible with pylint than the general projects pool)

[jacobtylerwalls]

Isn't keras failing in the "new" primer a valid reason to have it in the primer ? (If we only add projects that work from the get go we're selecting projects that are somewhat more compatible with pylint than the general projects pool)

Keras works locally when fewer projects are involved. We're blowing out available memory on Github Actions, I assume. We probably have too many projects in the primer anyway. If we ever parallelized the primer job, we could consider adding it back.

[Pierre-Sassoulas]

👌

[Pierre-Sassoulas]

I was going to squash but it occurred to me that keeping the 4 commit separated in the history could be preferable, but there's a conflict so it's impossible to rebase. I'll let you choose what to do :)

[jacobtylerwalls]

We can squash it; we don't need a super-fine resolution on a git blame for this testing tool...

[DanielNoord]

I think this regressed? See #8638, primer can't run anymore.

[jacobtylerwalls]

Taking a look