Add Vulkan backend to llama demo binary (#2608)

Summary:
Pull Request resolved: https://github.com/pytorch/executorch/pull/2608

## Context

Cleans up Vulkan backend CMake build so that `vulkan_backend` target is a single static library will all required symbols. Then add vulkan backend to llama runner binary.

imported-using-ghimport

Test Plan: Imported from OSS

Reviewed By: mcr229, cbilgin

Differential Revision: D55272794

Pulled By: SS-JIA

fbshipit-source-id: 16074d32315a01df0a2fd70311ae98c07da12dc9