(readgram): Use `#if 0' around code that accepted %command
inside grammar rules: The documentation doesn't allow it,
and it will fail since the %command processors scan for the next %.
(parse_token_decl): Extended the %token
declaration to allow a multi-character symbol as an alias.
(parse_thong_decl): New function.
(read_declarations): Added %thong declarations.
(read_declarations): Handle NOOP to deal with allowing
% declarations as another means to specify the flags.
(readgram): Allow %prec prior to semantics embedded in a rule.
(skip_to_char, read_declarations, copy_definition)
(parse_token_decl, parse_start_decl, parse_type_decl)
(parse_assoc_decl, parse_union_decl, parse_expect_decl)
(get_type_name, copy_guard, copy_action, readgram)
(get_type, packsymbols): Revised most error messages.
Changed `fatal' to `warnxxx' to avoid aborting for error.
Revised and use multiple warnxxx functions to avoid using VARARGS1.
(read_declarations): Improve the error message for
an invalid character. Do not abort.
(read_declarations, copy_guard, copy_action): Use
printable_version to avoid unprintable characters in printed output.
(parse_expect_decl): Error if argument to %expect exceeds 10 digits.
(parse_token_decl, parse_assoc_decl, parse_type_decl, get_type):
Allow the type of a non-terminal can be given
more than once, as long as all specifications give the same type.
