fix: Replace MPI barrier with an error collection system

[drodarie]

Describe the work done

• During boot, functions loading files may get stuck in parallel as one of the process is throwing an error but the others remain stuck at a MPI.Barrier(). Instead, this PR replaces the barrier with a status collector at the end of these functions to make sure each process is aware if one its fellow process is failing and raises an exception accordingly.
• Morphologies collected from NeuroMorpho will now be removed from the Scaffold if the NeuroMorphoSelector node is unset (through the __unboot__ function)

List which issues this resolves:

Fix #211, #197, #187

Tasks

• Added tests

---

📚 Documentation preview 📚: https://bsb-core--212.org.readthedocs.build/en/212/

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 85.57%. Comparing base (3a28863) to head (413a0b4).
⚠️ Report is 2 commits behind head on main.

Additional details and impacted files

@@            Coverage Diff             @@
##             main     #212      +/-   ##
==========================================
+ Coverage   78.67%   85.57%   +6.90%     
==========================================
  Files         196       86     -110     
  Lines       18146    11040    -7106     
  Branches     2138     1312     -826     
==========================================
- Hits        14277     9448    -4829     
+ Misses       3314     1300    -2014     
+ Partials      555      292     -263     

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.
• 📦 JS Bundle Analysis: Save yourself from yourself by tracking and limiting bundle sizes in JS merges.

[drodarie]

@Helveg, I know you suggested me to create utility functions in mpi.py that would implement the collect error system but I do feel that it might be difficult to generalize to every use case.

[Helveg]

ok is there nothing like:

with mpi.all_try():
  try:
    ..
  except Exception as e:
    errr.wrap(e, ...)

that could work in this case? If an error is raised the __exit__ call of the context manager receives the exception, and can allgather/broadcast it, and all nodes can raise it?

Maybe call them try_all and try_main?

[Helveg]

Probably cleaner to use the stdlib decorators

[Helveg]

LGTM, you choose what to do with the finally :)