Enable setting max tokens

Add llama3-generate

Use cache friendly KV Cache

Simplify the nextTick trick

Give GC a chance to run

Update node-mlx to 0.0.6

Publish via github workflow

Implement chat interface