fix test_ll for new llguidance with working advance_parser()

[mmoskal]

I have fixed the LLInterpreter.post_process() method to be the same as LLInterpreter.advance_parser() - they now actually correctly compute the new forced tokens after sampled token (before they would just return the sampled token and only next call to LLInterpreter.mid_process() would return the forced tokens).

The change above broke "my" test_ll.py - this fixed it (the test works now with both new and old llguidance). It didn't seem to break anything else (@hudson-ai told me that the new behavior was what he expected anyways).

One day I have to move this test to llguidance repo, but it's hard since I don't have any ergonomic way of producing the llguidance grammars there.

cc @Harsha-Nori

[codecov-commenter]

⚠️ Please install the to ensure uploads and comments are reliably processed by Codecov.

Codecov Report

All modified and coverable lines are covered by tests ✅

Project coverage is 61.36%. Comparing base (cf9ecd1) to head (8f8bce8).

❗ Your organization needs to install the Codecov GitHub app to enable full functionality.

❗ There is a different number of reports uploaded between BASE (cf9ecd1) and HEAD (8f8bce8). Click for more details.

HEAD has 4 uploads less than BASE

Flag	BASE (cf9ecd1)	HEAD (8f8bce8)
	84	80

Additional details and impacted files

@@            Coverage Diff             @@
##             main    #1013      +/-   ##
==========================================
- Coverage   70.14%   61.36%   -8.79%     
==========================================
  Files          62       62              
  Lines        4441     4441              
==========================================
- Hits         3115     2725     -390     
- Misses       1326     1716     +390     

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

[hudson-ai]

I can simplify a few things in parser.py now that I'm assured that post-process will advance us all the way (will do in a separate PR). Thanks @mmoskal !