feat: Add cat.starts_with/cat.ends_with

[mcrumiller]

Follow up to #20211.

Edit 2024-12-18: expression implementation has been removed, along with the slow path. cat.starts_with and cat.ends_with now require str inputs and do not except expressions.

[codecov]

Codecov Report

Attention: Patch coverage is 96.77419% with 2 lines in your changes missing coverage. Please review.

Project coverage is 79.10%. Comparing base (117a0ba) to head (c23630d).
Report is 42 commits behind head on main.

Files with missing lines	Patch %	Lines
crates/polars-plan/src/dsl/function_expr/cat.rs	93.54%	2 Missing ⚠️

Additional details and impacted files

@@            Coverage Diff             @@
##             main   #20257      +/-   ##
==========================================
- Coverage   79.65%   79.10%   -0.56%     
==========================================
  Files        1565     1572       +7     
  Lines      218281   220015    +1734     
  Branches     2475     2467       -8     
==========================================
+ Hits       173878   174046     +168     
- Misses      43836    45401    +1565     
- Partials      567      568       +1     

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

[ritchie46]

I am really not sure about the cast. Will have to think about this for a bit.

[mcrumiller]

I am really not sure about the cast. Will have to think about this for a bit.

We could simply disallow nonscalar expression inputs. cat.starts_with is a distinct namespace from str.starts_with so I think it's reasonable to have different rules in place regarding the allowable inputs. It would definitely simplify things as well.

Having nonscalar expression inputs does lean towards the "categoricals as string optimizations" (and in this case, it wouldn't even be an optimization) which I know is not their intended behavior. So--I propose only allowing scalar inputs here. This would completely remove the slow path. What do you think?

[mcrumiller]

@Ritchie I've removed the slow path, and switched to only allow str input arguments (similar to something like, say, str.strptime). I think this makes a lot more sense for categoricals. Since we no longer accept expressions, we have no more slow path. This also means that None is no longer accepted as well.