list.set_* on Categoricals/Enums gives confusing error messages

[shenker]

Checks

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

Reproducible example

Seen in polars-u64-idx 0.20.5 and 0.20.6-rc.1. Related: #11730, #11735.

Passing in a python list of strings to .list.set_* does not do what I'd expect (automatically cast argument to pl.List(pl.Categorical), to match the type of the Expr it's operating on). Furthermore, the error messages are confusing, making it difficult to figure out what's going on. If you explicitly pass dtype=pl.List(pl.Categorical) to pl.lit, it works as expected.

with pl.StringCache():
    df = pl.DataFrame(
        {
            "a": [
                ["x", "y", "z"],
                ["y", "z"],
                ["y", "z", "x1"],
                ["y", "z", "x2"],
                ["y", "z", "x3"],
            ]
        },
        schema=dict(a=pl.List(pl.Categorical)),
    )
    df.with_columns(pl.col("a").list.set_intersection(["x", "y"]).alias("b"))

prints

InvalidOperationError: casting to a non-enum variant with rev map is not supported for the user

with pl.StringCache():
    df = pl.DataFrame(
        {
            "a": [
                ["x", "y", "z"],
                ["y", "z"],
                ["y", "z", "x1"],
                ["y", "z", "x2"],
                ["y", "z", "x3"],
            ]
        },
        schema=dict(a=pl.List(pl.Categorical)),
    )
    df.with_columns(pl.col("a").list.set_intersection(pl.lit(["x", "y"], dtype=pl.Categorical)).alias("b"))

prints

PanicException: called `Result::unwrap()` on an `Err` value: ComputeError(ErrString("validity mask length must match the number of values"))

This works:

with pl.StringCache():
    df = pl.DataFrame(
        {
            "a": [
                ["x", "y", "z"],
                ["y", "z"],
                ["y", "z", "x1"],
                ["y", "z", "x2"],
                ["y", "z", "x3"],
            ]
        },
        schema=dict(a=pl.List(pl.Categorical)),
    )
    print(df.with_columns(pl.col("a").list.set_intersection(pl.lit(["x", "y", "x1"], dtype=pl.List(pl.Categorical))).alias("b")))

prints

┌──────────────────┬─────────────┐
│ a                ┆ b           │
│ ---              ┆ ---         │
│ list[cat]        ┆ list[cat]   │
╞══════════════════╪═════════════╡
│ ["x", "y", "z"]  ┆ ["x", "y"]  │
│ ["y", "z"]       ┆ ["y"]       │
│ ["y", "z", "x1"] ┆ ["y", "x1"] │
│ ["y", "z", "x2"] ┆ ["y"]       │
│ ["y", "z", "x3"] ┆ ["y"]       │
└──────────────────┴─────────────┘

Log output

No response

Issue description

n/a

Expected behavior

n/a

Installed versions

polars_u64_idx-0.20.6rc1