[core] Support bucket pruning for manifest files

[jianguotian]

Purpose

Bucket-key point lookups spanning many partitions can spend most of their planning time opening manifest files whose bucket ranges cannot match. This PR extracts only manifest bucket pruning from #9784, following the review feedback.

Changes

• persist one nullable totalBuckets value in manifest-list metadata
• record it only when every ADD and DELETE entry has the same positive total bucket count
• keep the value unknown for legacy manifests, copied unknown metadata, mixed counts, and non-positive counts
• reuse BucketSelector cached bucket sets to test whether candidate buckets intersect a manifest bucket range
• conservatively keep manifests when metadata or a usable predicate is missing, or when negative buckets are present

Manifest sorting, bucket-first layout, forced rewrites, and other independent optimizations are intentionally excluded and can be handled in follow-up PRs.

Verification

• focused manifest and bucket-selector tests: 106 tests, 0 failures, 0 errors
• paimon-core reactor tests: 5,389 tests, 0 failures, 0 errors, 34 skipped (PostgresqlCatalogTest excluded locally because Docker is unavailable)
• Checkstyle, Spotless, and Maven Enforcer: BUILD SUCCESS

[JingsongLi]

+1