REF: Remove PET estimator

[jhlegarreta]

Remove the PET estimator: the NiFreeze estimator design is modality-agnostic, so a PET-specific estimator is not required.

Remove the PET-specific integration test: checking that motion estimation with PET data yields the identity matrix when monkey patching the transformation computation when the estimator runs the registration step is covered by the
test_estimator:test_estimator_iterator_index_match test function.

Update the notebook accordingly:

• Avoid trying to fit on all data, as it is not implemented yet
  following commit a468d76.
• Call .matrix on the Affine instances to get the affine transformation as a NumPy array.

[jhlegarreta]

Supersedes PR #203.

Depends on PR #424.

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 83.37%. Comparing base (db0286b) to head (7007918).
⚠️ Report is 2 commits behind head on main.

Additional details and impacted files

@@            Coverage Diff             @@
##             main     #426      +/-   ##
==========================================
- Coverage   83.91%   83.37%   -0.54%     
==========================================
  Files          37       37              
  Lines        2138     2093      -45     
  Branches      235      231       -4     
==========================================
- Hits         1794     1745      -49     
- Misses        303      309       +6     
+ Partials       41       39       -2     

☔ 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.

[jhlegarreta]

@oesteban @mnoergaard Runs from beginning to end now. The estimation is still not good, but that is independent of the patch set I'd say.

[oesteban]

This is great. We can now work on making the general estimator work properly, but having a net decrease of ~2400 lines is a substantial complexity reduction. Thanks!

[jhlegarreta]

One thing that came to my mind yesterday was nipy/nitransforms#286 (comment). Maybe we are not extracting the motion parameters correctly, and hence, our results do not seem to be correct. May not be the only reason.