This introduces an alternative way of running AOT tests using the reference system added in apache#8514. This gives us additional assurance that the AOT output runs successfully on embedded platforms in our core test suite.

I've also changed calculate_workspace_sizes to debug_workspace_sizes and default to False in most cases as it only needs to be True for a few cases to check theoutput with the debug flag - this was discovered trying to allocate 16MB in an embedded test 🙀

Co-authored-by: Grant Watson <grant.watson@arm.com>