fix(python): Update implementation of Enum support in lit to address spurious test failure

[stinodego]

This is a bit of a weird one - I saw spurious failures on this Enum functionality which shouldn't have any randomness at all, e.g.:
https://github.com/pola-rs/polars/actions/runs/9661447486/job/26649165757?pr=17164

Hoping this new implementation takes care of it.

[codecov]

Codecov Report

All modified and coverable lines are covered by tests ✅

Project coverage is 80.82%. Comparing base (7a1be3b) to head (d6b4c64).

Additional details and impacted files

@@            Coverage Diff             @@
##             main   #17187      +/-   ##
==========================================
- Coverage   80.84%   80.82%   -0.03%     
==========================================
  Files        1464     1464              
  Lines      192016   192016              
  Branches     2743     2744       +1     
==========================================
- Hits       155243   155195      -48     
- Misses      36262    36310      +48     
  Partials      511      511              

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

[ritchie46]

Should we ensure we pass strings?

[stinodego]

Should we ensure we pass strings?

We already do... But Enum categories subclass string which makes them weird. I still have no clue how this could fail only sometimes... but new implementation should be better since it explicitly extracts the Enum value.