Update update_placeholder_tensor_specs to include lifted tensor constants (#4029)

Summary:
Pull Request resolved: https://github.com/pytorch/executorch/pull/4029

In `update_placeholder_tensor_specs` we're currently missing checking in `inputs_to_lifted_tensor_constants` to determine whether or not a placeholder is a constant. If this check is not performed, `spec.const` remains False, leading to the inclusion of this buffer size in the memory allocation for non_const_buffers, which will essentially force the runtime to allocate unneeded memory.

Reviewed By: mergennachin, JacobSzwejbka

Differential Revision: D58712012

fbshipit-source-id: f0bd87dbc4501c133b33b83a04910a8f1f57dbca