[FEA] Evaluate bloom filter impact for hash-based join operations

**Is your feature request related to a problem? Please describe.**
Hash-based joins (`hash_join`, `filtered_join`) probe the hash table for every input row. Bloom filters could eliminate non-matching rows before probes, but setup/query overhead may outweigh benefits. Need to evaluate speedup vs. penalty across different scenarios.

**Describe the solution you'd like**
Benchmark bloom filter prefiltering for hash-based joins:

- Measure performance with/without bloom filter for different selectivities (low/high match rates; various table sizes and key distributions and different bloom filter configurations
- Is it possible to find heuristics for enabling bloom filter?

**Additional context**
Can it be used to further improve mark join #21434?