Create Cold Start Performance benchmark

[bd82]

• Can we skip most of the grammar checks given a flag?
• Implications of grammar serialization on cold start performance?
• Could the lookahead functions be generated in advance?

[bd82]

• Could we serialize some of the results related to Lexer optimizations?
  • See: Performance improvement for large number of regex bd82/regexp-to-ast#17

[bd82]

We do not need anything fancy, Just add a similar code snippet
like this: https://github.com/SAP/chevrotain/pull/995/files#diff-82f87be4123976abd596f21427a33109
to the lexer init phase during the last version vs Master benchmark