[FEA] Improve performance of Java pinned memory pool at scale

**Is your feature request related to a problem? Please describe.**
We have seen cases where allocating many thousands of pinned memory buffers causes very poor performance in the Java `PinnedMemoryPool` allocator. That allocator currently performs a linear coalesce scan on free which performs decently when there aren't many allocations but does not scale well.

**Describe the solution you'd like**
The pinned memory pool should minimally use a logarithmic, heap-like algorithm for managing the address space, or possibly a bucketing technique with hashing.