LOAD_SUPER_ATTR doesn't match monitoring behavior of the opcodes it replaces

[carljm]

The LOAD_SUPER_ATTR super-instruction optimization doesn't correctly match the PEP 669 monitoring events that would be fired by the call to super() followed by attribute load that it replaces. This should be fixed.

Linked PRs

• gh-103865: add monitoring support to LOAD_SUPER_ATTR #103866