[flink] Lake-batch fallback should use KV snapshot instead of reading log from earliest

### Search before asking

- [x] I searched in the [issues](https://github.com/apache/fluss/issues) and found nothing similar.

### Description

Follow-up to #3296.

In batch mode on a lake-enabled PK table with no lake snapshot yet, the fallback reads every bucket's log from EARLIEST, ignoring existing Fluss KV snapshots. This is OOM-prone on large never-tiered tables.

Spark fixed the equivalent in #3317 with per-bucket dispatch (snapshot+tail where KV snapshot exists, log-only otherwise).

Flink can't 1:1 port because `FlinkSourceSplitReader` has no sort-merge on the Fluss-only path. We may reuse `LakeSnapshotAndLogSplitScanner` (already used by Flink's lake reader, already sort-merges per PK) for the fallback or do smth similar

### Willingness to contribute

- [ ] I'm willing to submit a PR!

Start by inspecting FlinkSourceSplitReader and the existing LakeSnapshotAndLogSplitScanner used by the Flink lake reader. Compare the fallback described here with the per-bucket Spark fix in #3317 and the preceding context in #3296. Done means batch fallback uses available KV snapshots with only the remaining log tail, while buckets without snapshots retain log-only behavior.

Written by the indexing model from the issue text.