bench: Add comprehensive multi-threaded benchmarks

Add concurrent benchmarks for all major operations to measure
multi-threaded scalability (1, 2, 4, 8, 16 threads):

- concurrent/claim: disjoint and contention scenarios
- concurrent/bitmap: set, test_and_set with contention
- concurrent/partition: partitioned iteration throughput
- concurrent/write_iter: concurrent claiming (empty and partitioned)
- concurrent/tracker: add, exists, mixed 80/20 read/write
- concurrent/group: registration, lookup, write round-robin
- concurrent/hierarchical: add_pair, claim_pair

These benchmarks help identify:
- Lock-free operation scaling
- Contention impact on CAS operations
- Partitioned iteration throughput
- Real-world multi-worker performance

Signed-off-by: Zvi Schneider <[email protected]>

Author: zvi-code