Measure the cost of per-instruction callback in bxcpu #141
Description
In #137, edge-coverage is implemented by using a set of hooks that executes on branches vs after every instruction. This means that potentially, we could stop using the per-instruction hook when fuzzing and potentially get a good performance win.
This task is about trying to find out how much of a speed-up this would bring, and if it looks decent to make edge coverage default, use per-instruction hook when tracing and implement a breakpoint mechanism that doesn't rely on the per-instruction hook.