remove support for UTF-16 and UTF-32

[Herringway]

UTF-16 and UTF-32 are so rarely used that the YAML devs are considering removing them from the spec entirely. For those who DO need this functionality, std.utf.toUTF8/toUTF16/toUTF32 covers it just as efficiently - we were eagerly converting under the hood anyways.

Notable public interface changes here:

• Loader.fromBuffer is now a deprecated alias of Loader.fromString as they are effectively equivalent now.
  • This means void[] or ubyte[] cannot directly be loaded anymore. I've only seen one public codebase even using this functionality, and it had to cast strings to ubyte[] to do it.
• Reader.encoding removed. Unsure if this was ever used.
• Token.encoding removed. ditto.
• toUTF8() was removed. Why was this public in the first place...?
• Emitter no longer has a CharType template parameter (only supports UTF-8)

Pros:

• We can drop our only dependency
• Simplified reading/writing code
• Less document copying (not part of this PR)
• Small but noticeable reduction to compile time

Cons:

• Breaks backwards compatibility, but hopefully not in a way that actually impacts anyone

[Herringway]

After nearly a year with no concerns raised, it seems like a good time to merge and see what happens.