Handle invalid expressions

[sunmy2019]

A new test case failure can be added.

f"{6 0}"

The old parser will try to send "6 0" into the parser for f-string, thus will raise an SyntaxError like

    (6 0)
     ^^^
SyntaxError: f-string: invalid syntax. Perhaps you forgot a comma?

The new parser will (force) except } right after 6. Thus the error message would be

    f"{6 0}"
         ^
SyntaxError: expected '}'

The new parser cannot match the invalid_* at the first round, but p->error_indicator was set to 1, thus it cannot perform the second round.

https://github.com/python/cpython/blob/0708437ad043657f992cb985fd5c37e1ac052f93/Parser/pegen.c#L835-L840

So I made some changes to the grammar file to avoid setting p->error_indicator in that case.

---

As for the original test case,

Even with this change, the new parser will output

    f"{6 0="
           ^
SyntaxError: f-string: expecting '}'

rather than the old parser's

    (6 0)
     ^^^
SyntaxError: f-string: invalid syntax. Perhaps you forgot a comma?

After some digging, this error was improperly hijacked by the tokenizer:
https://github.com/python/cpython/blob/0708437ad043657f992cb985fd5c37e1ac052f93/Parser/pegen_errors.c#L406-L416

I think we should just leave it here, it is acceptable.

[pablogsal]

This PR needs now rebase