Fix GANN indexing and VC dimension

Delegate GANN load and search to C library

Add C code for GANN traversal

Batch inserts, parallelize GANN tree loading

Add MongoDB sink and consolidate Milvus interface

Flatten GANN tree structure for efficient traversal

Update project description

Optimize GANN queries with Numba

Add UUIDs to GANN nodes

Add sampling for large datasets in GKMeans

Allow GANN indexes to be stored and loaded

Explicitly use nodes for GANN tree traversal

Use hardmax for GKMeans forward pass

Slight bugfixes and optimizations

Update default LR, avoid unnecessary CPU-GPU transfers