Add stats for "unlikely events" that will impact optimization.

[markshannon]

Events such as using changing the __code__ attribute of a function, or the __bases__ of a class, are rare.
At least we presume that they are rare.
We want to optimize code with the assumption that they are rare (see See faster-cpython/ideas#645).

So, we needs stats to confirm that they are rare, and hopefully give us some insight into how rare.

We should add a struct to the interpreter to track these events as they occur, and if stats are enabled add that data to the stats output.

Linked PRs

• gh-114312: Collect stats for unlikely events #114493
• gh-114312: Fix rare event counter tests on aarch64 #114554

[markshannon]

We should track the following events:

• Setting an object's class, obj.__class__ = ...
• Setting the bases or mro of a class, cls.__bases__ = ...
• Setting the PEP 523 frame eval function, _PyInterpreterState_SetEvalFrameFunc().
• Modifying the builtins, __builtins__.__dict__[var] = ...
• Function modification, setting the defaults, closure, etc. func.__closure__ = (None, None)

These two events are also mentioned in faster-cpython/ideas#645, but we don't need stats for them (for now)

• Modification of locals in a debugger. Not a property of the program, so stats would be meaningless.
• Instrumentation, sys.monitoring.set_events(). We already have plenty of stats for this.

For each event we want a _Py_stats_struct counter, as normal.
We also want counters available to the optimizer, so each event should also get a saturating uint8_t counter attached to the interpreter.