feat(python,rust): Improve usability of Null type.

[ghuls]

Improve usability of pl.Null type:

• Creating a Null array with repeat in eager works now: s = pl.repeat(None, 7, eager=True) Also bounds checks for converting to i32 series are slightly tweaked for lazy.
• Converting Polars DataFrames with Null Series to PyArrow tables works now. pl.DataFrame([pl.Series("null", [None, None], dtype=pl.Null)]).to_arrow()

[ghuls]

Some other things I noticed:

• Formatting of dataframe columns with pl.List(pl.Null) does not work properly.
• A plain Polars Null series does not get printed like other series.

>>> tbl = pa.table({"a": [None, None], "b": [[None, None], [None, None]]})
>>> pl.from_arrow(tbl)
shape: (2, 2)
┌──────┬─────────────────┐
│ a    ┆ b               │
│ ---  ┆ ---             │
│ null ┆ list[null]      │
╞══════╪═════════════════╡
│ null ┆ fmt implemented │
│ null ┆ fmt implemented │
└──────┴─────────────────┘
>>> df = pl.from_arrow(tbl)
>>> df["a"]
nullarray

pl.repeat does not work exactly the same for eager and lazy. Lazy can return pl.Int32 or pl.Int64 depending if the value fits in int32 boundaries for numbers while eager always returns pl.Int64.
@ritchie46 is this intentional?

[ritchie46]

Nope, this is not intentional. Can you make an issue? Yeap, I think we should implement the fmt for the nullarray as well.

[ghuls]

I made an issue for the fmt problem: #7153