Dont memory plan for inputs (#2155)

Summary:
Pull Request resolved: https://github.com/pytorch/executorch/pull/2155

For KV cache with IO tHis results in
1. allocating kv cache in the memory plan but also allocated by llama runner
2. Doing actual copy of kv cache

Also we should really make plan_input = false by default. I dont imagine a case
where this does not result in making copies. Planning for output is fine but
still dangerous as people may assume having reference to output tensor is all
good without realizing the underlying memory being shared.
ghstack-source-id: 216889056
exported-using-ghexport

validated oss ci is clean. have to by pass because ci think its needs internal linter to pass.
bypass-github-export-checks

Reviewed By: mergennachin

Differential Revision: D54161288

fbshipit-source-id: b5e7aa42d4a72e455550af5d7467f46f2a1017f8