[FEAT] any_value groupby aggregation

[kevinzwang]

This function is parameterized by ignore_nulls, which attempts to find a non-null value in each group when true. However, usage of this parameter in the aggregation function would require some changes to DataFrame._agg() that I am going to save for later, since these changes will probably not be needed anymore once global expressions can be passed into GroupBy operations

Also in this PR: fixes to the count aggregation function

[codecov]

Codecov Report

Attention: Patch coverage is 80.00000% with 1 lines in your changes are missing coverage. Please review.

Project coverage is 83.93%. Comparing base (9c66a5e) to head (c5bfb4f).
Report is 10 commits behind head on main.

Additional details and impacted files

@@            Coverage Diff             @@
##             main    #1941      +/-   ##
==========================================
- Coverage   85.40%   83.93%   -1.47%     
==========================================
  Files          55       55              
  Lines        6221     6132      -89     
==========================================
- Hits         5313     5147     -166     
- Misses        908      985      +77     

Files	Coverage Δ
daft/expressions/expressions.py	91.39% <100.00%> (+0.41%)	⬆️
daft/logical/builder.py	89.16% <ø> (-1.00%)	⬇️
daft/dataframe/dataframe.py	87.15% <50.00%> (-0.13%)	⬇️

... and 15 files with indirect coverage changes

[samster25]

Zooming out a bit, we shouldn't have to implement this for each for the Arrays separately since this operation only has to work on the validity mask.
You could actually implement it on the series directly.

let mask = self.validity();
let indices = get_idx_from_bitmap(mask, groups);
self.take(indices)

I think count could also be refactored that way.
What do you think?

[kevinzwang]

I implemented this idea for any_value, but the array-level count implementation is used directly in parts of the code (eg. for computing mean) so I think it's still useful to consolidate the logic there