fix: suppress nested progress bars via thread-local context manager

[BHARATH0153]

closes #1811

changes

Add sbi/utils/pbar.py with thread-local nesting tracker. Outer samplers (accept_reject_sample, rejection_sample, sampling_importance_resampling) wrap proposal calls in nested_pbar_context, while downstream functions like Diffuser.run check is_nested() to suppress their own bars.

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 87.99%. Comparing base (dbf432b) to head (f7cf2b0).
⚠️ Report is 21 commits behind head on main.
✅ All tests successful. No failed tests found.

Additional details and impacted files

@@            Coverage Diff             @@
##             main    #1922      +/-   ##
==========================================
+ Coverage   87.88%   87.99%   +0.11%     
==========================================
  Files         143      144       +1     
  Lines       13387    13677     +290     
==========================================
+ Hits        11765    12035     +270     
- Misses       1622     1642      +20     

Flag	Coverage Δ
fast	81.33% <100.00%> (?)

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

Files with missing lines	Coverage Δ
sbi/samplers/importance/sir.py	100.00% <100.00%> (ø)
sbi/samplers/rejection/rejection.py	92.62% <100.00%> (+0.18%)	⬆️
sbi/samplers/score/diffuser.py	91.93% <100.00%> (+0.13%)	⬆️
sbi/utils/__init__.py	100.00% <100.00%> (ø)
sbi/utils/pbar.py	100.00% <100.00%> (ø)

... and 3 files with indirect coverage changes

[BHARATH0153]

@janfb please review when you have time thanks!

[BHARATH0153]

@Jocho-Smith please review your insight will be helpful thanks!

[Jocho-Smith]

Hi @BHARATH0153 , I like the slim and precise implementation. How exactly did you test that this is leading to the desired output?

I just run this test here again: #1473 (comment)

Instead of the the current

Generating 10 posterior samples in 4 diffusion steps  
Generating 100 posterior samples in 4 diffusion steps  
66it [00:01, ...]

I see with your code

(sbi) schmidt@node-63:~/gsoc2026/sbi$ pytest -s tests/test_npse_auto_gauss_iid.py
66it [00:01, 45.03it/s] 

So no bars at all.

This might be one possible solution to resolve the confusion, however I think this is not what you aimed for here. So the confusion from the test example from the link above is that one requests 10 samples and gets 2 status bars mentioning 10 and 100 additional samples. This can be confusing, b/c users might think they got 100, when they actually get the 10 requested.

So your solution removes the bars entirely, which makes the nestedness tracking unnecessary, right?

Maybe it is helpful to look at the example code I linked and design your solution it in a way that makes it behave it in the way you want - possibly writing some more 'test scenarios'. Does that make sense? Hope it helps.

Quick note: I'll be busy the next weeks as well, so Jan might actually be faster reviewing it before I can reply the next time. Thanks for looking at it either way!

[BHARATH0153]

@janfb please check the new changes when you are free thanks!

[janfb]

Hi @BHARATH0153 please note that as said in my comment below the issue that I am not able to review this in the next weeks. I will try to review it early August. Thanks for your patience!