This contains the same fix as #1901, but for ocamllex actions. Without this, actions like

rule token = parse
  | 'a' { f' '"' }
  | 'b' { f2 '\o170' '"' }
  | 'c' { f1 "\u{1F42B}" }
  | 'd' { f1 {|}|} }
  | 'e' { (* " *) } (* " *) }

would not compile, while they contain valid ocaml code.

The changes are

• recognize identifiers containing single quotes
• recognize octal character literals
• recognize unicode character escape sequences
• recognize quoted strings

I tried to keep it a bugfix, not a language change. But as a side effect, octal and unicode escape sequences are now allowed in ocamllex strings too.