Preserve SDPA for CoreML (#5258)

Summary:
## Motivation
Starting from iOS18, CoreML has added SDPA op, so there is no longer need to decompose torch SDPA

## Solution
Following https://github.com/pytorch/executorch/pull/3483, add `ops_not_to_decompose` in CoreML partitioner, then use `to_edge_transform_and_lower` API in llama export

Pull Request resolved: https://github.com/pytorch/executorch/pull/5258

Reviewed By: kirklandsign

Differential Revision: D62550916

Pulled By: cccclai

fbshipit-source-id: bc238a04ea9eb38341157cd2388b7cd17a506bf1