[core][python] Improve earliest snapshot retry handling

[JingsongLi]

Summary

When the earliest snapshot is expired concurrently, callers may observe an earliest snapshot id whose snapshot file is already gone. This PR makes Java and Python retry forward through a bounded window before giving up, while preserving empty-table handling for callers that have no available latest snapshot.

Changes

• Share Java earliest-snapshot retry handling through SnapshotManager.retryEarliestSnapshot, reused by both SnapshotManager and TimeTravelUtil.
• Align Python SnapshotManager with the Java behavior by retrying from the discovered earliest snapshot id instead of stopping at the first missing snapshot.
• Increase the default earliest-snapshot retry window from 3 to 300 snapshot ids to better tolerate batch expiration races.
• Throw RuntimeException / RuntimeError when an identified earliest snapshot range cannot be read after retry, instead of silently returning null / None.
• Remove repeated WARN logs from the Java retry loop.
• Add regression coverage for missing earliest snapshots, retry exhaustion, and retrying past more than one hundred missing Python snapshots.

Testing

• GitHub Actions checks for this PR passed.
• python -m pytest paimon-python/pypaimon/tests/snapshot_manager_test.py paimon-python/pypaimon/tests/time_travel_util_test.py
• git diff --check