Enable Partial GPU lowering via Vulkan in stories model export (#2368)

Summary:
Pull Request resolved: https://github.com/pytorch/executorch/pull/2368

## Context

Simple change to add Vulkan Partitioner as a dependency for the llama exporter and runner, and provide a command line flag to invoke the vulkan partitioner during export.

ghstack-source-id: 218708315
exported-using-ghexport

bypass-github-export-checks

Reviewed By: digantdesai, kirklandsign

Differential Revision: D54805831

fbshipit-source-id: 7c2b5072254afb41a173b94ed72bbbe42a0832fc