Arrow: Support Large Binary when using to_arrow

[castedice]

This PR is to address an issue that prevented the to_arrow method from handling binaries larger than 2GB when used as mentioned in #344.

With this change in place, all binary types must be defined via pa.large_binary when defining a pyarrow schema.

I considered leaving it as pa.binary and casting it in pyiceberg, but since defining it as pa.large_binary is essential for pyarrow to handle 2GB of data, I implemented it this way.

[Fokko]

@castedice Thanks for raising this. I think this is fine since Polars does it as well:

python3
Python 3.11.7 (main, Dec  4 2023, 18:10:11) [Clang 15.0.0 (clang-1500.1.0.2.5)] on darwin
Type "help", "copyright", "credits" or "license" for more information.
>>> import polars as pl
>>> df = pl.DataFrame(
...     {"foo": [1, 2, 3, 4, 5, 6], "bar": [b"a", b"b", b"c", b"d", b"e", b"f"]}
... )
>>> df.to_arrow()
pyarrow.Table
foo: int64
bar: large_binary
----
foo: [[1,2,3,4,5,6]]
bar: [[61,62,63,64,65,66]]

[castedice]

Thanks for review
This change will require a few changes to the documentation.
After checking the documentation, I'll create an additional PR.