Make the parser conform to ICU MessageFormat

[nkovacs]

Apostrophes are now handled correctly, emulating
ICU's default DOUBLE_OPTIONAL behavior.

Octothorpe is handled correctly, and can be escaped using
apostrophes, but only inside plural
(depending on strictNumberSign).

A single apostrophe only starts quoted literal
text if it immediately precedes a curly brace ({}),
or, if inside a plural, an octothorpe (#).
The parser now supports the strictNumberSign option,
since that determines whether a quoted octothorpe
is parsed as '#' or just #.

Since choice format isn't supported, the pipe symbol
never causes an apostrophe to start quoted literal text.

Parameters to functions may contain whitespace and
quoted special characters, but argStyle is still trimmed
and split into multiple parameters.
A new option, strictFunctionParams, activates ICU-compatible
parsing, which parses everything from the second
comma to the closing curly brace as a single "argStyleText"
parameter.

Fixes #1, fixes #2

[jsf-clabot]

All committers have signed the CLA.

[eemeli]

Sorry for taking so long to get to this. Overall, I like these changes. However, it would be good if you could split the quote-escaping changes into a separate PR, as that's a change that'll require us to update the major version of the parser -- it will break some users' strings.

[eemeli]

Why not return p.trim()?

[nkovacs]

Because _ is just [ \t\n\r] and trim is [\s], which is [ \f\n\r\t\v\u00a0\u1680\u2000-\u200a\u2028\u2029\u202f\u205f\u3000\ufeff], and I wanted to keep it the same.

Changing them both to \s would probably be fine.

[eemeli]

Yah, let's do that. The _ definition is an ancient remnant.

[nkovacs]

I've split the function parameter stuff into #4.
It should be backwards-compatible.

[nkovacs]

\s doesn't work in peg.js. It has to be done manually: nkovacs@a0be6aa

Also, the identifier syntax in the Java implementation is [^[[:Pattern_Syntax:][:Pattern_White_Space:]]]+, where Pattern_Syntax is this: http://unicode.org/cldr/utility/list-unicodeset.jsp?a=%5B:Pattern_Syntax:%5D&abb=on&g= and Pattern_White_Space is http://unicode.org/cldr/utility/list-unicodeset.jsp?a=%5B%3APattern_White_Space%3A%5D&abb=on&g=&i=
See http://icu-project.org/apiref/icu4j/com/ibm/icu/text/MessageFormat.html, http://icu-project.org/apiref/icu4j/com/ibm/icu/text/SelectFormat.html and http://icu-project.org/apiref/icu4j/com/ibm/icu/text/PluralFormat.html

The id definition in messageformat-parser is a bit inconsistent. The first character has to be ascii alphanumeric or $ or _, but then the rest of the characters can be almost anything.