using COCOINDEX_CODE_DEVICE=mps can enhance embedding speed if you are apple silicon mac user

[Slone123c]

This should be informed in the readme. Otherwise the default behavior is using cpu to embedding.

[badmonster0]

@Slone123c appreciate the suggestion! cc @georgeh0

[georgeh0]

@Slone123c thanks for the suggestion.

This part was over engineered. Created #50 to simply automatically detect device when the environment variable is not set, so MPS will be chosen on macOS automatically.