[FIX] force_fmapless on top of self-correcting pepolar bold

[bpinsard]

When force_fmapless is used on top of self-correcting pepolar bold, the fmapless estimator will get it's name from B0FieldIdentifier, creating a duplicate.
For now I just created a test to replicate that behavior and triggering the error that this PR will try to fix.

[bpinsard]

Also, it seems there was no test for force_fmapless=True.

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 83.75%. Comparing base (f7abb18) to head (efa1e97).
⚠️ Report is 129 commits behind head on main.

Additional details and impacted files

@@           Coverage Diff           @@
##             main     #465   +/-   ##
=======================================
  Coverage   83.75%   83.75%           
=======================================
  Files          30       30           
  Lines        2826     2826           
  Branches      367      367           
=======================================
  Hits         2367     2367           
  Misses        387      387           
  Partials       72       72           

☔ View full report in Codecov by Harness.
📢 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.

[bpinsard]

Not sure how that will behave with multi-subject cases, seee #455 .
Maybe that should be covered in tests.

[effigies]

I'm not sure how I feel about this, now that I've implemented application of pre-computed fieldmaps. In practice, the way to discover the relevant fieldmaps was to re-run the SDCFlows fieldmap wrangling function, and then look for precomputed fieldmaps of the relevant IDs.

This will change that ID, making a hard break between pre- and post-change derivatives. One of the ostensible advantages to the fit-transform split was that we can run an improved transform stage on any previous derivatives, so long as the derivatives are queryable.

We would need to strategize an alternative method for finding the appropriate fieldmaps from both new and old derivatives to make this change.

[bpinsard]

I am not sure to fully grasp the challenges here, and I wiped a lot of my memory of why I submitted that PR.

• the BIDS-id prefix that I added would likely break BIDS naming, but that can easily be fixed by something only alphanumerical
• do you worry that reexecution of the wrangler in the apply run would generate IDs different from the fit one, and thus fail to retrieve existing fieldmap. If so, how can we fix the autonaming in FieldmapEstimation to avoid the collision observed above that caused crashes?