fix: Fix use of COUNT(*) in SQL GROUP BY operations

[alexander-beedie]

The difference between len/count caused the aggregate expression check in "GROUP BY" to miss len (it's not actually an AggExpr), as used by "COUNT(*)" in the SQL interface. Fixed, and added the missing coverage.

(Also includes two minor unrelated docstring updates).

Example

df = pl.DataFrame({
    "a": ["xx", "yy", "xx", "yy", "xx", "zz"],
    "b": [1, None, 3, 4, None, 6],
    "c": [99, 99, 66, 66, 66, 66],
})

df.sql("""
  SELECT
    a,
    COUNT(b) AS count_b,
    COUNT(*) AS count_star
  FROM self
  GROUP BY a
  ORDER BY a
""")

# shape: (3, 3)
# ┌─────┬─────────┬────────────┐
# │ a   ┆ count_b ┆ count_star │
# │ --- ┆ ---     ┆ ---        │
# │ str ┆ u32     ┆ u32        │
# ╞═════╪═════════╪════════════╡
# │ xx  ┆ 2       ┆ 3          │
# │ yy  ┆ 1       ┆ 2          │
# │ zz  ┆ 1       ┆ 1          │
# └─────┴─────────┴────────────┘

[codecov]

Codecov Report

All modified and coverable lines are covered by tests ✅

Project coverage is 81.45%. Comparing base (d5f9c3b) to head (e611fa2).
Report is 9 commits behind head on main.

Additional details and impacted files

@@            Coverage Diff             @@
##             main   #16465      +/-   ##
==========================================
+ Coverage   81.42%   81.45%   +0.02%     
==========================================
  Files        1409     1409              
  Lines      184618   184726     +108     
  Branches     2971     2971              
==========================================
+ Hits       150334   150474     +140     
+ Misses      33764    33732      -32     
  Partials      520      520              

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