[python] Push down shard range to Vortex and Lance format readers

[chenghuichen]

Purpose

Previously, SlicedSplit row ranges were applied post-read via ShardBatchReader — the full file was read before discarding out-of-range rows.

This change pushes the shard range directly into format readers that support native range reads:

• Lance: uses LanceFileReader.read_range(start, num_rows) so only the requested rows are read from disk
• Vortex: converts the shard range to indices for VortexFile.scan(indices=...), leveraging native row selection

Other formats (Parquet/ORC/Avro/Blob) fall back to ShardBatchReader as before.

Tests

• paimon-python/pypaimon/tests/reader_append_only_test.py