Add support for MPS in the basehandler

[agunapal]

🚀 The feature

Based on a config in model-config.yaml, ex: mps:enable, Basehandler should move the model to mps https://pytorch.org/docs/stable/notes/mps.html

Motivation, pitch

To make use of the GPU on apple silicon mac

Alternatives

CPU support for apple silicon already exists

Additional context

No response

[lxning]

The ticket requires:

1. a model can be deployed to either CPU or GPU on Mac based on model-config.yaml
2. a mac can have 1+ GPU. a model can be deployed to a specified GPU device id on Mac.