Added support for UnicodeCombiningMark, fixes #3639.

[ctjlewis]

Some notes:

1. Annotating functions with spec grammar is likely wise (and will help with issues like this in the future), so I did not remove that JSDOC block.
2. A utils/ directory was added, which contains a script for building the source (without GWT) and a script for calling that built version with the given args. This directory will be a good place to put bash scripts in the future, including runtime_tests.sh to execute the jsComp runtime tests (a separate issue I'm working on).
3. There is more grammar to implement if we want to be consistent with the spec (isConnectorPunctuation, isZeroWidthJoiner, etc.), which will prevent issues like this in the future. I intend to do this over the coming days, I just wanted to get this PR in first.

This PR fixes #3639, and you can see that the source:

var bar = {
  İ: "foo"
};

now compiles as expected to:

var bar={"I\u0307":"foo"};

[ctjlewis]

Build tests are failing because:

/src/com/google/javascript/jscomp/parsing/parser/Scanner.java:909: error: cannot find symbol
    return Character.getType(ch) == Character.NON_SPACING_MARK;
                    ^
  symbol:   method getType(char)
  location: class Character
/src/com/google/javascript/jscomp/parsing/parser/Scanner.java:909: error: cannot find symbol
    return Character.getType(ch) == Character.NON_SPACING_MARK;

Anybody have recommendations? Not sure why Character.getType would not be defined, I even added a manual import for java.lang.Character, and getType method has been defined since JDK 1.1.

I noticed the rest of the character checks did not use builtins like Character.NO_SPACING_MARK, but this syntax is much more legible on top of being future-proofed and likely just as performant.

Edit: Is this breaking because it's being fed into J2CL, which can't compile Character.getType and therefore doesn't import it?

[ctjlewis]

Opened an issue at google/j2cl#103 regarding future-proofed Unicode category checks, just need J2CL to support Character.getType() and the category constants.

[ctjlewis]

Running ./build_test.sh created bazel output directories that were added to the git tree by default. These lines prevent that.

[ctjlewis]

Not compliant (see #3647), per spec UnicodeCombiningMark also includes "Non-spacing mark (Mn)" category, but this addresses #3639 and a decent scope of similar possibilities with as minimal change as possible.

[ctjlewis]

Handy for debugging local build.