Use-after-free is detected in the core interpreter. How would Miri add its span information to the use-after-free diagnostics?