Performance improvements

[curlpipe]

I could potentially look at a concurrent highlighting solution, where single line tokens, multiline tokens and bounded tokens are all calculated in parallel, hopefully that would speed up performance significantly.

[curlpipe]

For v2 of synoptic, this could be implemented as firing up multiple threads and passing them different lines in the document, of which they can then search for atoms to create. Then, after atoms are found, tokenization can occur.