[spark] lake table predicate pushdown

### Search before asking

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


### Description

Forward converted Fluss predicates to `LakeSource.withFilters(...)` for both lake-enabled append (log) and upsert (primary-key) batch reads, matching the Flink behavior.

### Willingness to contribute

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