[wpilib, commands] Cache controller BooleanEvents/Triggers and directly construct Triggers

[Gold856]

Fixes #5903, resolves #6445 by updating it to work with the HID templates. (Updating those is great by the way, you just have to change a few lines and regenerate!)

[github-actions]

This PR modifies commands. Please open a corresponding PR in Python Commands and include a link to this PR.

[Gold856]

I don't think I need to do anything for Python? Python Commands already directly constructs Triggers, and since objects are ref-counted, I don't think there's much of a GC concern either.

[KangarooKoala]

I'm concerned about the axis caches, since there's no guarantee that the threshold is in the range [0, 50). (First, it's plausible that someone would want a negative threshold for e.g. the left x axis (which would lead to a conflict between the axisGreaterThan and axisLessThan caches), and second, although it wouldn't make sense, a user could pass a really high number, which should be handled in some way to avoid having a really difficult to track bug.)

[Gold856]

Yeah, I'm not sure how to handle axis caches because they're done in a unique way. I suppose I could nest the Map even further (Map<EventLoop, Map<Integer, Map<Double, Trigger>>>), but I feel like that's starting to just be extra overhead for a feature that people might not use.

[Alextopher]

I try to lean towards keeping maps as flat a possible. How about Map<EventLoop, Map<Pair<Integer, Double>, Trigger>. Then there is no chance a custom key function can be wrong.

[PeterJohnson]

The C++ classes need mutexes as well.

[Gold856]

C++ classes don't cache BooleanEvents or Triggers, so they don't need mutexes.