The existing behavior of bdb's monitoring backend does not fully match whatsys.settrace does. The update foropcode events is separated from the other local events, which does not quite make sense.

This PR unifies the event handling -update_local_events() will simply checkf_trace andf_trace_opcode to decide what events should be enabled for that code object.

This actually cleans up the logic for the monitoring backend.

Thecheck_trace_opcodes() is removed fromstop_trace() becauseclear_id does the work.

There was a small bug in the original code -check_trace_opcodes will actually erase the workcheck_trace_func does. It did not matter because we always dostart_trace() when we start trace and we are only interested in opcode events. However, theoretically it's still wrong. Now it has the correct behavior.

This is not really "user observable", so it should be considered as a refactoring. (However it blocks something I'm working on for async stuff).

• Issue:Make it possible to usesys.monitoring for pdb/bdb #120144