Refactor PlatformManager for performance

[anonyco]

Two improvements sure to provide a nice speedup to WorldEdit:

1. Desynchronize <PlatformManager>.queryCapability, which is used in some pretty tight loops that would feel the benefits of less synchronized.
2. Remove the use of EnumMap from PlatformManager.preferences, which further optimizes <PlatformManager>.queryCapability because EnumMaps do incur some overhead, mostly from checking to make sure the enum you passed it is really of the same class as the enum of the EnumMap.

[me4502]

Refactors for performance that make the code substantially messier (the fact you didn't follow the style guide at all aside), without any proof of speed-up or evidence that it's necessary, won't be accepted.

Provide a microbenchmark and CPU profile that show why these changes make sense to do if you want to PR something like this.

I'm also not convinced it couldn't be done in a substantially cleaner manner, if it's true that this has a performance impact

[anonyco]

I have significantly simplified these changes to the bare minimum: the removal of just the synchronized.

1. I was unaware that there was a style guide. I read all the markdown and txt files, including CONTRIBUTING.md. Could you refer me to where to find the style guide?

2. The results of my microbenchmark are below:

Before:
     Calculated BlockType.getRichName x10000 in 84128654ns
     Calculated BlockType.getRichName x10000 in 14269327ns
     Calculated BlockType.getRichName x10000 in 7806132ns

After:
     Calculated BlockType.getRichName x10000 in 60178043ns
     Calculated BlockType.getRichName x10000 in 7953745ns
     Calculated BlockType.getRichName x10000 in 4852244ns

Across the runs above, we see 40%-79% better performance for BlockType.getRichName with this small change.

3. I am very unfamiliar with the available Java tools and do not know "CPU profile." I am not asking you to waste your time explaining every step to me, rather I would really appreciate it if you could give me a piece of text or a name of a software/tool that I could look up and research to learn more about what the common way to do this is in the WorldEdit project.

Many thanks for your time and attention to this.

[octylFractal]

I would rather refactor the whole class to stop using synchronized at all, and instead take advantage of a ReadWriteLock (probably ReentrantReadWriteLock to start with). It'll increase performance everywhere and is not as potentially dangerous as unguarded reads from a thread-unsafe map. If you don't feel up to that, please close this PR and I will file an issue to do so at a later date.

[me4502]

closing in favour of #2571