Add a Rudimentary Core ML Partitioner (#1415)

Summary:
In this PR, we add a rudimentary Core ML partitioner based on `CapabilityBasedPartitioner`, which proposes possible partitions based on whether a PyTorch op can run on the specified backend, in our case, Core ML.

Any insights on potential ways to implement the Core ML partitioner are welcomed :pray:

The unit test has to be run with [this coremltools branch](https://github.com/apple/coremltools/pull/2081)

Pull Request resolved: https://github.com/pytorch/executorch/pull/1415

Reviewed By: digantdesai

Differential Revision: D52791944

Pulled By: shoumikhin

fbshipit-source-id: d49b3249f2c686bc38a2379227836874b16001e1