Only 2 free reserved token slots remained in gen_lex_token.cc.

The current patch:

* forces explicit token numbers in %token declarations in sql_yacc.yy and
  sql_hints.yy

* extends the token space with extra segment for new tokens

* cleanups the code

* fixes query normalization bugs:

  -- OR2_SYM should output "||" to the normalized query string instead of "|"

  -- token NEG is not related to "~"

Note: the fix doesn't renumber previously existent tokens, so, digests
test results should not alter.

Change-Id: Ic47ba4359e18403e7f5edd1a7d510c416d2cc736