fix: Properly broadcast list arithmetic

[cmdlineluser]

ref: #18831

It seems like I just needed to copy the broadcast_array() impl from #18851

[codecov]

Codecov Report

Attention: Patch coverage is 86.95652% with 3 lines in your changes missing coverage. Please review.

Project coverage is 79.89%. Comparing base (262f7bc) to head (bc8029e).
Report is 1 commits behind head on main.

Files with missing lines	Patch %	Lines
...polars-core/src/series/arithmetic/list_borrowed.rs	86.95%	3 Missing ⚠️

Additional details and impacted files

@@            Coverage Diff             @@
##             main   #18858      +/-   ##
==========================================
+ Coverage   79.87%   79.89%   +0.02%     
==========================================
  Files        1519     1519              
  Lines      205839   205852      +13     
  Branches     2898     2898              
==========================================
+ Hits       164416   164475      +59     
+ Misses      40875    40829      -46     
  Partials      548      548              

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

[itamarst]

I still think some of these tests are necessary? E.g. pl.Series([[1, 2], [3, 4]]) + pl.Series([[1, 1], [2, 2], [3, 4]]) should complain.

(Separately I am honestly not super-excited about semantics of single-item Series working this way, but I guess that's how literals work? So if that design decision has already been made, oh well.)

[itamarst]

Or rather, some variations of existing tests should still be there, by tweaking existing assertions instead of deleting them.

[itamarst]

You should also test eager operations in addition to lazy operations, e.g. df.get_column("A") + lit, it sometimes hits different code paths for each when doing arithmetic, so you get different bugs.

I would also suggest adding a test where the literal is a different type than the list, e.g. a Int64, and again testing both lazy and eager (casting definitely works differently).

[cmdlineluser]

Thank you @itamarst

I made a naive copy of the broadcast_array() PR as I thought it would be in time for 1.8 to have it working for both array and list.

I will close this as you are working this area already.

[itamarst]

I'm not working on the same thing though, so other than merge conflicts this seems worth continuing.

[itamarst]

But if you don't want to finish it, I can finish it up.