Improve parser performance when handling special characters

[changhiskhan]

thousands separator and comments (#962) processing is slow right now.

at least comments should be preprocessed

[wesm]

Thousands handling is handled at a low level in the new parser. Have to think about the comments argument and how to do it without slowing down the tokenizer

[wesm]

Thousands handling is way faster, but the C parser doesn't do comments yet. Should be simple (ish) to add to the tokenizer.