- mark actual EBNF with pre-formatted class "ebnf" instead of "grammar"

- make real productions for Unicode char classes so that they can be parsed
- use `` for tokens that contain "'s or \'s so that they can be parsed
- added a missing '.'

This version of the spec passes through ebnflint (forthcoming) without errors.

R=r,rsc
DELTA=74  (3 added, 1 deleted, 70 changed)
OCL=31464
CL=31466