match_query: unit test and rewrite for TypeScript

[liamwhite]

This moves all of the match_query clientside parsing library into TypeScript. There should be no substantial logic changes to the internal parse logic here, but three things were discovered during testing and fixed:

1. Incorrect escaped quote processing in terms. This arose from the translation of the following Ruby code

   def normalize_term!(match, quoted)
     if quoted
       match.gsub!('\"', '"')
     else
       match.gsub!(/\\(.)/, '\1')
     end
   end

   as

   SearchTerm.prototype._normalizeTerm = function() {
     if (!this.wildcardable) {
       return this.term.replace('\"', '"');
     }
     return this.term.replace(/\\([^\*\?])/g, '$1');
   };

   which is a very subtle mistake, as the string literal escaping rules are different in JS.
2. Inconsistency in relative date parsing. For absolute dates, the lte and gt operators use the "upper" date in the date range, and the lt and gte operators use the "lower" date. For relative dates, only the "lower" date was used. This is inconsistent with the behavior of the server-side parser, which uses the same behavior for both date types, so it was changed to match.
3. Use of the deprecated/optional substr JS API. These correspond to the slice method used in the Ruby parser.

   # Truncate string and restart the token tests.
   @search_str = @search_str.slice(match.size,
                                   @search_str.size - match.size)

   Just a minor fix to change these to substring.

This has 100% test coverage in unit tests and has also been manually integration tested.