Refactoring LL(*) feature

[msujew]

Refactors major parts of the LL(*) feature, adding new features:

• [x] Made ATNSimulator into a parser trait
• [x] "Fixed" failing tests, adding a few new ones that test specific LL(*) behavior
  • I've skipped the ECMA5 "quirk" tests for now. They have to be rewritten some way or another to accommodate the LL(*) feature
  • Achieved 100% code coverage (not 100% branch coverage ... yet) of lookahead code
• [x] Added runtime ambiguity reporting
• [x] Removed static ambiguity analysis (validation)
• [x] Removed lots of dead code

Food for thought

Keep the old static ambiguity analysis for LL(k) decisions, where k < maxLookahead. For example, such an ambiguity can be easily identified statically:

OR([{
  DEF: () => this.CONSUME1(A) 
}, {
  DEF: () => this.CONSUME2(A) 
}])