fix: Match runtime behavior of (Table|RecordBatch).select (#221)

* fix: Match runtime behavior of `(Table|RecordBatch).select`

## Resolves
- https://github.com/MarcoGorelli/narwhals/blob/5b02b592183b8d39e2d32e0aedd6c234bb22d405/narwhals/_arrow/dataframe.py#L305-L307
- https://github.com/MarcoGorelli/narwhals/blob/5b02b592183b8d39e2d32e0aedd6c234bb22d405/narwhals/_arrow/dataframe.py#L285-L294

##Description
Following up on what I thought was a simple stub issue, but we're both *too strict* and *too permissive* in different ways
##Examples
{placeholder}

##Related
- https://github.com/apache/arrow/blob/d2ddee62329eb711572b4d71d6380673d7f7edd1/python/pyarrow/table.pxi#L4367-L4374
- https://github.com/apache/arrow/blob/d2ddee62329eb711572b4d71d6380673d7f7edd1/python/pyarrow/table.pxi#L1721-L1739

* update select

* update select

---------

Co-authored-by: ZhengYu, Xu <zen-xu@outlook.com>

Author: dangotbanned

pyarrow-stubs/__lib_pxi/table.pyi

@@ -28,6 +28,7 @@ from typing import (
 import numpy as np
 import pandas as pd
 
+from numpy.typing import NDArray
 from pyarrow._compute import (
     CastOptions,
     CountOptions,
@@ -525,7 +526,7 @@ class RecordBatch(_Tabular[Array]):
     def serialize(self, memory_pool: MemoryPool | None = None) -> Buffer: ...
     def slice(self, offset: int = 0, length: int | None = None) -> Self: ...
     def equals(self, other: Self, check_metadata: bool = False) -> bool: ...
-    def select(self, columns: list[str] | Indices) -> Self: ...
+    def select(self, columns: Iterable[str] | Iterable[int] | NDArray[np.str_]) -> Self: ...
     def cast(
         self, target_schema: Schema, safe: bool | None = None, options: CastOptions | None = None
     ) -> Self: ...
@@ -592,7 +593,7 @@ JoinType: TypeAlias = Literal[
 class Table(_Tabular[ChunkedArray[Any]]):
     def validate(self, *, full=False) -> None: ...
     def slice(self, offset=0, length=None) -> Self: ...
-    def select(self, columns: list[str] | Indices) -> Self: ...
+    def select(self, columns: Iterable[str] | Iterable[int] | NDArray[np.str_]) -> Self: ...
     def replace_schema_metadata(self, metadata: dict | None = None) -> Self: ...
     def flatten(self, memory_pool: MemoryPool | None = None) -> Self: ...
     def combine_chunks(self, memory_pool: MemoryPool | None = None) -> Self: ...