Improve Lookahead functions and maxLookahead handling

[bd82]

Changes

• Moved Lookahead Function computation to parser initialization.
  - This Increases the initialization time, but gained 5-15% runtime performance boost (Parser only benchmark, no Lexer involved).
• Support maxLookahead config per DSL method rather than just a global setting.

  class LowLookaheadParser extends CstParser {
      constructor() {
          super([], {
              // Globally **only one** token lookahead.
              maxLookahead: 1
          })
  
          $.RULE("value", () => {
              $.OR({
                  // We need **two** tokens lookahead to distinguish between these two alternatives
                  MAX_LOOKAHEAD: 2,
                  DEF: [
                      {
                          ALT: () => {
                              $.CONSUME(A)
                              $.CONSUME(B)
                          }
                      },
                      {
                          ALT: () => {
                              $.CONSUME(A)
                              $.CONSUME(C)
                          }
                      }
                  ]
              })
          })
  
          this.performSelfAnalysis()
      }
  }

---

Original Issue

The LA functions are evaluated lazily when needed.
This is useful when parsing small inputs that may not need all of the LA functions.

However if by creating these all LA functions at startup it may be possible to:

• Improve runtime speed by avoiding the check if the LA function has been created.
• Improve analysis speed by memoization of previous results.
• Enable inspecting the actually needed maxLookahead versus the one defined by the user.
• Enable providing these lookahead functions "pre-prepared" e.g via some kind of code generation. this could both:
  • Reduce init/first run time.
  • Increase runtime performance by providing more optimized lookahead functions

[bd82]

WIP here: #1022