Preserve "garbage"

[ExplodingCabbage]

I've decided this may have been misguided. It's way too complicated and the correct behaviours to implement are not clear. Several dilemmas:

• Should we also support parsing/preserving trailing garbage? I can't find any programs that output it, yet the GNU patch util supports ignoring it. What's the right behaviour?

• What about trailing garbage after a hunk header, on the same line as it? Git heuristically generates such garbage to try to show the enclosing context (e.g. function declaration) that a hunk appears in - e.g. the export function formatPatch(diff) { here:

  diff --git a/src/patch/create.js b/src/patch/create.js
  index e04f13d..462a171 100644
  --- a/src/patch/create.js
  +++ b/src/patch/create.js
  @@ -165,10 +165,6 @@ export function formatPatch(diff) {
      ret.push.apply(ret, hunk.lines);
    }
  
  -  if (diff.trailingGarbage) {
  -    ret.push(diff.trailingGarbage);
  -  }
  -
    return ret.join('\n') + '\n';
  }

  if we don't support that, we're still not really letting people roundtrip their patches through a parsePatch / formatPatch combo.

• What should formatPatch do if someone passes it a structured patch object where they've manually set the leadingGarbage to include file headers - i.e. to not be valid garbage? We should probably reject that somehow, but that further complicated the API.

Not worth it - at least not without a lot more subtle and nuanced design and implementation work than I've already put into this PR. If someone wants to take that on in due course, good for them, but I don't want to merge something half-assed because I fear I'm just making a mess that's worse than the status quo.

I will carefully cherry-pick out the actual bugfix from this PR, but I'm abandoning the rest.