WIP: Decouple game and ui frame rates

[ngc92]

This is some work in progress towards decoupling the frame rate of the ui and the game simulation, and also with the goal of getting rid of the monstrosity half-singleton with function-level statics that is the current SpeedController.

It is still unfinished, single player mostly works but I haven't tested network or replays.
Still, I hope this can serve as a foundation to discuss and design what we want to achieve.

Some observations:

1. Do we still need the show-fps option? On desktops, you almost certainly should be able to achieve the target fps nowadays, and on mobile you won't really see the title bar change in any case.
2. Ultimately, I would like to move game and ui to different threads. Even cheap mobile devices nowadays have multiple cores, so that should help further ensure good performance.
3. How to do the waiting. Right now I'm using std::this_thread::yield(), but that seems to generate still 100% CPU on linux.
4. Increase granularity of game simulation. Instead of 75 steps per second, e.g. 150. This allows more precise input (high-end desktop with 140 Hz monitor might want to run input at the same frequency) and makes the blobby positions seem less discrete (e.g. more variety when serving). Should not be problematic from a computation point of view, currently I can run about 20k game steps per second on a single thread.
5. Instead of spawning on thread per game on the server, have a pre-existing thread pool (configurable) and assign new games to these threads. With non-blocking speed control, we can have games with different speeds on the same thread.

Thoughs?

[ngc92]

So, point 4 is a bit more involved, as the bot scripts have parts that use hard-coded numerical constants that may depend on details of the game simulation. I still think it is a good change, but we have to go over the bot scripts and make adjustments accordingly. Ideally also getting rid of as many fixed constants as possible.
Some of the things I've noticed when looking over the bot scripts, which are independent of this change, I've put into #101