Add bucket-transform-aware partition pruning

[guillesd]

Equality and IN-list predicates on bucket()-partitioned columns now prune files via ducklake_file_partition_value at scan-planning time. Bucket hashes are computed in C++ so the catalog SQL stays backend-agnostic. Range predicates fall through to the existing zone-map path.

Basically makes lookup queries useful with buckets. Zone-maps are not really effective for hashed transforms.

[guillesd]

We could alternatively make BuildBucketPartitionPruningClause part of GenerateFilterPushdownComponents and centralize everything there. I'd just need to change the signature of the function to get DuckLakeTableEntry to have the partition info. Let me know what you think @pdet

[pdet]

Thanks, LGTM, can you just fix the failing CI?

[guillesd]

@pdet done! Basically all partitioning tests are skipped in Quack in the first place because of a known limitation. So I just added the new test to the skip list only for Quack!

[pdet]

Thanks!