Cannot use `.list` namespace functions in GroupBy context

Checks

[X] I have checked that this issue has not already been reported.
[X] I have confirmed this bug exists on the latest version of Polars.

Reproducible example

import polars as pl

df = pl.DataFrame({
    "A": [1, 1, 2, 2],
    "B": [1, 2, 3, 4],
})
df.group_by("A").agg(pl.col("B").list.to_array(2))

Log output

polars.exceptions.ComputeError: expected List dtype

Error originated just after this operation:
DF ["A", "B"]; PROJECT */2 COLUMNS; SELECTION: "None"

Issue description

When aggregating a column in a GroupBy context, the list namespace is not accessible.

Expected behavior

Should be able to call .list namespace functions.

Installed versions

``` --------Version info--------- Polars: 0.20.8 Index type: UInt32 Platform: Linux-5.15.133.1-microsoft-standard-WSL2-x86_64-with-glibc2.35 Python: 3.12.1 (main, Jan 31 2024, 09:51:46) [GCC 11.4.0] ----Optional dependencies---- adbc_driver_manager: 0.9.0 cloudpickle: 3.0.0 connectorx: deltalake: 0.15.3 fsspec: 2023.12.2 gevent: 24.2.1 hvplot: 0.9.2 matplotlib: 3.8.2 numpy: 1.26.4 openpyxl: 3.1.2 pandas: 2.2.0 pyarrow: 15.0.0 pydantic: 2.6.1 pyiceberg: 0.5.1 pyxlsb: 1.0.10 sqlalchemy: 2.0.27 xlsx2csv: 0.8.1 xlsxwriter: 3.1.9 ```

yes, I was looking for a good solution here but I think there is a "problem" in polars in the agg context which can be solved in two ways:

more "hacks" (make it somehow work in egg context)
rethink the automatic list conversion if you use pl.col(...) which is a big problem imo

Auto `list` cast of `pl.col(...)` in agg context

This has often been discussed and I personally think this is a bad idea. This "bug" is just one example

The user should be responsible for bringing the data in the aggregation down to a single line. I think auto casting pl.col to a list is problematic in many ways, one example is this issue.

The problem is a confusing inconsistency between the select and agg context. It is very similar in many ways but then has some quirks like this which makes it confusing. Why not make them identical in terms of user experience?

example:

df = pl.DataFrame({"A": [1, 1, 1], "B": [1, 2, 4]})

Select

df.select(
    min=pl.col("B").min(),
    max=pl.col("B").max(),
    mean=pl.col("B").mean(),
    implode=pl.col("B").implode(),                         # explicit `implode` required!
    implode_array=pl.col("B").implode().list.to_array(3),  # ".list" namespace available
)
shape: (1, 5)
┌─────┬─────┬──────────┬───────────┬───────────────┐
│ min ┆ max ┆ mean     ┆ implode   ┆ implode_array │
│ --- ┆ --- ┆ ---      ┆ ---       ┆ ---           │
│ i64 ┆ i64 ┆ f64      ┆ list[i64] ┆ array[i64, 3] │
╞═════╪═════╪══════════╪═══════════╪═══════════════╡
│ 1   ┆ 4   ┆ 2.333333 ┆ [1, 2, 4] ┆ [1, 2, 4]     │
└─────┴─────┴──────────┴───────────┴───────────────┘

select requiers explicit implode to get a list type and make the ".list" namespace available

group_by / agg

df.group_by("A").agg(
    min=pl.col("B").min(),
    max=pl.col("B").max(),
    mean=pl.col("B").mean(),
    col=pl.col("B"),  # imo this should not work! No auto cast to list
    implode=pl.col("B").implode(),  # this should create a list (not list of list)
    implode_array=pl.col("B").implode().list.to_array(3),  # this should work like in select
)
shape: (1, 7)
┌─────┬─────┬─────┬──────────┬───────────┬─────────────────┬─────────────────────┐
│ A   ┆ min ┆ max ┆ mean     ┆ col       ┆ implode         ┆ implode_array       │
│ --- ┆ --- ┆ --- ┆ ---      ┆ ---       ┆ ---             ┆ ---                 │
│ i64 ┆ i64 ┆ i64 ┆ f64      ┆ list[i64] ┆ list[list[i64]] ┆ list[array[i64, 3]] │
╞═════╪═════╪═════╪══════════╪═══════════╪═════════════════╪═════════════════════╡
│ 1   ┆ 1   ┆ 4   ┆ 2.333333 ┆ [1, 2, 4] ┆ [[1, 2, 4]]     ┆ [[1, 2, 4]]         │
└─────┴─────┴─────┴──────────┴───────────┴─────────────────┴─────────────────────┘

agg works like select in almost everything but then has this weird auto cast to list type which causes problems as there is only a "magic-list-transformation" but not really that makes the ".list" namespace unavailable
imo this should work exactly like select in requiring explicit implodes which would be consistent and make the ".list" namespace available immediately

pola-rs / polars