Improve hash partition for columns contains Struct/list/map

[laysfire]

Description

This pr is to improve hash partition performance when table contains pandas can't handle types by moving retrieve table columns operations out of loop.
Use the following script to verify:

import time
import pyarrow as pa
from ray.data._internal.arrow_ops.transform_pyarrow import hash_partition

idx = list(range(50000000))
ints = [[i]for i in range(50000000)]
t = pa.Table.from_pydict(
   {
       "idx": pa.array(idx),
       "ints": pa.array(ints),
   }
)

start = time.time()
hash_partition(t, hash_cols=["idx", "ints"], num_partitions=10)
end = time.time()
print(end - start)

The test result is:

CPU spec	Code Version	Time consumed
Apple M4	original	66s
Apple M4	optimized	18s

Related issues

Link related issues: "Closes #62550", Cant' reopen after force push, recreate new pr.

[gemini-code-assist]

Code Review

This pull request optimizes the row-by-row hashing logic in _hash_partition by replacing manual column indexing with a more efficient enumerate(zip(*table.columns)) approach. This change improves code readability and potentially performance when processing unhashable PyArrow columns. I have no feedback to provide as there are no review comments.

[laysfire]

@owenowenisme hi, could you help review?

[owenowenisme]

Great improvement! Could you add comments explaining this improvement in the code?