refactor: remove deprecated x_shape where not needed.

[janfb]

x_shape as argument to posteriors is deprecated. We still have for backwards compatibility in some posteriors.

However, we don't need it in the new posteriors. We can also remove it in some internal uses - which is done in this PR.

[codecov]

Codecov Report

All modified and coverable lines are covered by tests ✅

Project coverage is 78.27%. Comparing base (8afd985) to head (3c6d680).
Report is 11 commits behind head on main.

Additional details and impacted files

@@            Coverage Diff             @@
##             main    #1271      +/-   ##
==========================================
- Coverage   86.05%   78.27%   -7.79%     
==========================================
  Files         118      119       +1     
  Lines        8672     8695      +23     
==========================================
- Hits         7463     6806     -657     
- Misses       1209     1889     +680     

Flag	Coverage Δ
unittests	78.27% <ø> (-7.79%)	⬇️

Flags with carried forward coverage won't be shown. Click here to find out more.

Files with missing lines	Coverage Δ
sbi/inference/posteriors/ensemble_posterior.py	50.00% <ø> (-37.97%)	⬇️
sbi/inference/trainers/fmpe/fmpe.py	94.28% <ø> (-0.06%)	⬇️
sbi/inference/trainers/npse/npse.py	96.45% <ø> (-0.03%)	⬇️
sbi/inference/trainers/nre/nre_base.py	94.44% <ø> (-0.06%)	⬇️

... and 29 files with indirect coverage changes

[gmoss13]

Thanks @janfb! I think it's a good idea, but was hoping you could clarify why we remove the x_shape argument from, e.g. NRE or ensemble_posterior, which are not new?

[janfb]

Thanks @janfb! I think it's a good idea, but was hoping you could clarify why we remove the x_shape argument from, e.g. NRE or ensemble_posterior, which are not new?

I agree, we should not remove it from the ensemble posterior init.

I think it's fine to remove from NRE because it is used there only internally, i.e., it is set but then not used.

[janfb]

This is passed as None explicitly here, i.e., the x_shape argument was not added to the user-facing constructor in the first place.
Thus I think it's fine to remove it here.

[janfb]

self._x_shape is not used internally anymore, so it's safe to remove it here.

[janfb]

@gmoss13 I added explanations in comments. What do you think?

[gmoss13]

Okay, thanks for clarifying! I agree with your comments