enh(parser) beginKeyword no longer bestows double relevance

[joshgoebel]

Previously something like:

{ 
  beginKeywords: "def"
  // which also implies:
  // keywords: "def"
}

Would bestow 2 relevance. 1 for the rule match and a second for the
keywords match. This is not correct behavior. A single "token" should
default to a single relevance point, not 2.

Changes

This changes the default relevance of a mode containing beginKeywords
to 0 forcing it to only gain relevance from keyword matches alone.

This of course revealed a few auto-detect balance issues (because some
grammars had been benefiting from this extra boost) which this PR
also resolves.

Checklist

• Added markup tests, or they don't apply here because...
• Updated the changelog at CHANGES.md
• Added myself to AUTHORS.txt, under Contributors

[joshgoebel]

Commit history here is clean if one wanted to review per commit.

[egor-rogov]

Looks good, but do we really need floating-point relevance (1.5)? I guess 2 didn't work, but still...

[joshgoebel]

Looks good, but do we really need floating-point relevance (1.5)? I guess 2 didn't work, but still...

Well, need is a strong word. I don't want to encourage it (if not necessary), but it was helpful here; and honestly DOUBLE relevance (not to mention 10x) is really a pretty huge multiple to be the only choice. And many of the really "smart" things we could do with relevance will ultimately result in fractional relevance values.

Math.floor is just me being ridiculously careful with compatibility. I wager it wouldn't matter, but was easy enough to add and then remove with v11 (allowing fractional relevance values).