feat: Add list.pad_start()

[etiennebacher]

Fixes #10283

Currently, this allows to do this:

import polars as pl

(
  pl.DataFrame(
    {
      "a": [[1], [], [1, 2, 3]], 
      "int": [0, 999, 2], 
      "float": [0.0, 999, 2]
    }
  )
  .select(
    filled_int=pl.col("a").list.pad_start(fill_value=pl.col("int")),
    filled_float=pl.col("a").list.pad_start(fill_value=pl.col("float"))
  )
)

shape: (3, 2)
┌─────────────────┬───────────────────────┐
│ filled_int      ┆ filled_float          │
│ ---             ┆ ---                   │
│ list[i64]       ┆ list[f64]             │
╞═════════════════╪═══════════════════════╡
│ [0, 0, 1]       ┆ [0.0, 0.0, 1.0]       │
│ [999, 999, 999] ┆ [999.0, 999.0, 999.0] │
│ [1, 2, 3]       ┆ [1.0, 2.0, 3.0]       │
└─────────────────┴───────────────────────┘

(
  pl.DataFrame({"a": [["a"], [], ["b", "c", "d"]]})
  .select(pl.col("a").list.pad_start(fill_value="foo"))
)

shape: (3, 1)
┌───────────────────────┐
│ a                     │
│ ---                   │
│ list[str]             │
╞═══════════════════════╡
│ ["foo", "foo", "a"]   │
│ ["foo", "foo", "foo"] │
│ ["b", "c", "d"]       │
└───────────────────────┘

Questions:

1. In the linked issue above, the workaround lets the user specify the final length of each sublist but I use the length of the longest sublist and pad all other sublist to match this length instead. Is this correct?
2. I feel like there's too much code duplication in the match statement, but it's not obvious to me how to reduce it. Any idea?
3. Are the failures in new-streaming expected?

[codecov]

Codecov Report

Attention: Patch coverage is 98.03922% with 2 lines in your changes missing coverage. Please review.

Project coverage is 79.76%. Comparing base (87baf86) to head (2d38618).
Report is 30 commits behind head on main.

Files with missing lines	Patch %	Lines
...tes/polars-ops/src/chunked_array/list/namespace.rs	98.73%	1 Missing ⚠️
crates/polars-plan/src/dsl/function_expr/list.rs	87.50%	1 Missing ⚠️

Additional details and impacted files

@@            Coverage Diff             @@
##             main   #20674      +/-   ##
==========================================
+ Coverage   78.85%   79.76%   +0.91%     
==========================================
  Files        1558     1561       +3     
  Lines      221055   221887     +832     
  Branches     2527     2530       +3     
==========================================
+ Hits       174318   176996    +2678     
+ Misses      46159    44309    -1850     
- Partials      578      582       +4     

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

[etiennebacher]

It should be possible to support more types (date, datetime, duration, maybe categorical and enum also?), but I'd like to have some feedback on the current implementation before, cf the two questions in the first post.

In particular, for now it casts categoricals to string in this case:

pl.DataFrame({"a": [["a"], ["a", "b"]]}, schema={"a": pl.List(pl.Categorical)}).select(
    pl.col("a").list.pad_start("foo")
)

# shape: (2, 1)
# ┌──────────────┐
# │ a            │
# │ ---          │
# │ list[str]    │
# ╞══════════════╡
# │ ["foo", "a"] │
# │ ["a", "b"]   │
# └──────────────┘

which I'm not sure is correct.