Merge: gamnit: use the same code to limit fps than mnit and fix errors

Fixed a few bugs in `mnit::mnit_fps`, a typo invalidated the value of `current_fps` and the use of long Ints were broken probably since they are tagged. Using floats should fix this last issue.

Finally, the goal of this PR is to copy the module `mnit::mnit_fps` to `gamnit::limit_fps` and adapt it to the gamnit framework. There is some code duplication, but the mnit version should be deleted with the engine when gamnit is completed.

Pull-Request: #1879
Reviewed-by: Jean Privat <jean@pryen.org>