LV2: Introduce Dynamic Timer signals

[elad335]

For a while, the emulator has dealt with LV2 timed syscalls pretty inefficiently:

• With "Usleep" setting: the emulator would keep the thread awake for the last 500 (on Windows, 100 for the rest) microseconds of the syscall duration awake, which had increased CPU time significantly while also not actually working as expected in a highly active threading environment due to forced operating systems preemptions.
• With "As Host" setting: none of the above, but the delay in the syscall schedule was huge! leading to various performance issues as well.

Instead, try a new thing: we know that direct thread notifications are very fast on every operaing systems, so try to systematically check for time-outs by other emulationh threads when they have little to no work themeselfs. This would eliminate the delay and cut on CPU time!
But hold on! what about times when other emulation threads do have work to be ?
The truth is that we cannot predict such cases with full accuracy, but what we can do is calculate the avarage response time of other threads to this in the past in games and see if it is valueable to opt for this mechanis instead. At the times when not, fallback to "Useleep" setting behaviour. This would provide accuracy in the "not well suited" cases too!

This adds the setting value "Dynamic" for LV2 timers accuracy, while it being the new default.

What to test:

1. PPU usage.
2. CPU usage.
3. Framerate
4. If Nier works properly.

Results: around 20% in CPU time specifically used by PPU in usleep syscalls is saved now, it's nice but it's a bit disappointing because there may be more potential there. That's it for this pr though.

[elad335]

It seems like the majority of USLEEP calls in games happen in loop and for short periods of time, while not currently targeted, I will add a case for it in the new setting.