Move towards "no ticks in CPU emulator" design

[dogtopus]

Having the check_stop_condition code hook slows down the emulation a lot (currently uses ~25% of the CPU time basically doing nothing). Since so far all the synchronization happens in timers and status register checks (instead of depending on hardware timing characteristics), we might be able to only keep a timer event scheduling thread (similar to ffsds4's EventSequencer) that asynchronously inject timer interrupt events into the emulator, and insert hooks into status register reads to either return a "not ready" when the emulator cannot or does not want to do a task synchronously, or "ready" after synchronously finishing the task.

[dogtopus]

Or we may not need the thread at all. We can add a new instruction hook to Unicorn that detects b pc-8 and stop the emulator. Though we also need to take different architecture type into account when doing this--armv4 and v5 should accept b pc-8 without issue, but what about v6 and v7+?

[dogtopus]

^ That seems to overcomplicate the problem a lot. Maybe going with a more straightforward route instead, like "inter-thread procedure call" over MPSC channels (device running outside CPU and using MPSC to communicate with the CPU thread when it needs data or need to send a signal, and vice-versa when CPU needs attention from a device).