Improving performances

[lowlighter]

Since #7, we use Flux (~ Deno.reader & Deno.seeker) to navigate through parsed stream/string.

Unfortunately there are some performance issues, which could maybe be improved by using some sort of cache.

input	v1	v2	v2 + progress logged in stdout
string	~7s/mb	~12s/mb	~18s/mb
stream	N/A	~42s/mb	~53s/mb

[phil294]

I am running some kind of batch script and for ~1000 parse iterations, this library took 2.8s while both https://deno.land/x/[email protected] and https://deno.land/x/[email protected] were at around .2s, so around 10 times faster. Hope this helps. I like this library though, thanks for your work.