Model Sharding

Shard large models across multiple nodes for distributed inference with repartir.

cargo run --example distributed_model_sharding