isScalar needs to reject invalid ASCII

[AZero13]

0x80 - 0xFF are invalid ascii, and thus, need to not return true always.

[glessard]

This function is an implementation detail of the function _UTFParser.parseScalar(from:), where the input is expected to have already been validated. It is used to implement a fast path where a single code unit from the input happens to be a whole scalar. For ASCII, this is always true, and we do not need to revalidate it. While there is no bug to fix here, in principle this change should be harmless. It is possible, however, that the change would slow things down a tiny bit. I would be interested in finding out if a misbehaviour caused by this function can be teased out?