[python] Fix pyarrow6 compatibility issue with path parsing and get_file_info

[XiaoHongbo-Hope]

Purpose

1. REST read and write fail with error:
   oss://bucket_name/xxxx/yyy/sample/bucket-0/test-0.parquet: When listing objects under key 'yyy/sample/bucket-0' in bucket 'xxxx': AWS Error [code 133]: The specified key does not exist.

PR #7180 addressed OSS + PyArrow 6 by passing key-only (no netloc), but PyArrow 6 still has a issue about parsing the first path segment as bucket.

Behavior difference:

• PyArrow 6: treats path a/b/c as bucket=a, key=b/c
• PyArrow 7+: uses bucket from connection and treats entire string as key a/b/c

2. get_file_info() inconsistency:
   • PyArrow 7+: returns FileInfo with FileType.NotFound for non-existent paths
   • PyArrow 6: throws OSError with message AWS Error [code 133]: The specified key does not exist

This PR will fix the above issue.

Tests

file_io_test.test_exists
file_io_test.test_filesystem_path_conversion

API and Format

Documentation

Generative AI tooling

co-authord by Cursor 2.4.28