[python] Avoid an extra file-size I/O for Blob reads

[XiaoHongbo-Hope]

Purpose

Avoid the extra file-size lookup performed when SplitRead creates a FormatBlobReader. DataFileMeta already contains the file size, so object-store reads do not need an additional metadata request before reading the Blob footer and index.

Changes

• Add an optional file_size argument to FormatBlobReader.
• Reuse DataFileMeta.file_size in both normal and Data Evolution raw Blob read paths.
• Preserve the existing lookup when FormatBlobReader is constructed without a valid file size.

Tests

• Provided file size avoids FileIO.get_file_size.
• Missing or zero file size falls back to one lookup.
• Both SplitRead Blob paths forward DataFileMeta.file_size.
• Blob tests: 186 passed, 1 unrelated Ray/pandas compatibility test deselected, 44 subtests passed.
• Deferred Blob tests: 15 passed.
• Flake8 and git diff --check passed.

[JingsongLi]

+1