[WIP] `lark`-based EBNF grammar builder, RFC 3986 test case

[hudson-ai]

trafficstars

The EBNF class is still pretty clunky, and I'm playing around with figuring out what the right ergonomics are. Might not want to call that part of the "public API" yet, but the functions exposed in _rfc3986.py should be fair game.

I'm not sure how tightly coupled I want us to be with lark -- maybe we should support various dialects of (G/E)BNF.

Note: test-suite success depends on #1016

[codecov-commenter]

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

Codecov Report

Attention: Patch coverage is 89.33333% with 8 lines in your changes missing coverage. Please review.

Project coverage is 61.75%. Comparing base (418fc03) to head (0c89e3f).

Files with missing lines	Patch %	Lines
guidance/library/_ebnf.py	89.28%	6 Missing :warning:
guidance/library/_rfc3986.py	89.47%	2 Missing :warning:

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

:exclamation: There is a different number of reports uploaded between BASE (418fc03) and HEAD (0c89e3f). Click for more details.

HEAD has 16 uploads less than BASE

Flag	BASE (418fc03)	HEAD (0c89e3f)
	22	6

Additional details and impacted files

@@            Coverage Diff             @@
##             main    #1017      +/-   ##
==========================================
- Coverage   69.32%   61.75%   -7.57%     
==========================================
  Files          62       64       +2     
  Lines        4443     4518      +75     
==========================================
- Hits         3080     2790     -290     
- Misses       1363     1728     +365     

:umbrella: View full report in Codecov by Sentry.
:loudspeaker: Have feedback on the report? Share it here.

[hudson-ai]

Closing as lark support is moving into llguidance