fix(gnn-node): Use Float32Array for NAPI bindings to fix type conversion errors

[ruvnet]

Summary

• Changed Vec<f64> parameters to Float32Array in all GNN node bindings
• Fixes "Failed to convert napi value Object into rust type f64" errors
• Aligns GNN bindings with the working pattern used in @ruvector/attention

Changes

Updated functions to use Float32Array:

• RuvectorLayer.forward(): now takes Float32Array parameters and returns Float32Array
• TensorCompress.compress(): now takes Float32Array embedding
• TensorCompress.compressWithLevel(): now takes Float32Array embedding
• TensorCompress.decompress(): now returns Float32Array
• differentiableSearch(): now takes Float32Array query and candidates
• hierarchicalForward(): now takes Float32Array query and layer_embeddings

Also updated JavaScript tests to use Float32Array.

Test plan

• Code compiles with cargo check -p ruvector-gnn-node
• Rust tests pass with cargo test -p ruvector-gnn-node
• JavaScript tests pass (requires native build)
• Integration test with AgentDB

Breaking Change Notice

This is a breaking change for users of @ruvector/gnn package. Users need to update their code to pass Float32Array instead of plain JavaScript arrays:

Before:

const query = [1.0, 2.0, 3.0, 4.0];
const result = layer.forward(query, neighbors, weights);

After:

const query = new Float32Array([1.0, 2.0, 3.0, 4.0]);
const result = layer.forward(query, neighbors, weights);

Fixes #35

🤖 Generated with Claude Code