modify making token by byte position, not rune position at readName

[nktks]

Hi!

I tried to parse SDL with multibyte comment , and failed.

I think, the cause is that the start and end positions of the token created in readName were created as the rune position.
If the comment contains a multibyte character string, the start and end of the token shifts from the actual byte position, and it seems that the token is not parsed correctly.

Please review this.

Thanks!

[coveralls]

Coverage remained the same at 92.422% when pulling ea0716e on nakatamixi:fix/multi-byte-parse into 8a92e97 on graphql-go:master.