from_arrow handles empty/duplicate column names badly

[stinodego]

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

import pyarrow as pa
import polars as pl

a1 = pa.array([1, 2])
a2 = pa.array([3, 4])


df = pl.from_arrow(pa.Table.from_arrays([a1, a2], names=["", ""]))
print(df.columns)  # ['column_0', 'column_1']

df = pl.from_arrow(pa.Table.from_arrays([a1, a2], names=["", "a"]))
print(df.columns)  # ['column_0', 'a']

df = pl.from_arrow(pa.Table.from_arrays([a1, a2], names=["a", "a"]))
print(df.columns)  # ['a']

Log output

No response

Issue description

pyarrow Tables allow duplicate column names, while we do not.

We handle this issue for empty columns by replacing them with column_0, column_1, etc.
For duplicate named columns, we simply drop any duplicates (only keeping the last one).

Expected behavior

I propose we raise an error on duplicate column names, with the suggestion to specify the schema argument. Then we can treat empty column names ("") the same as any other column name.

Installed versions

main branch, pyarrow 13.0.0