More Efficient extractDomElements

[dajinchu]

optimize extractDomElements to not construct a dom. instead we simply run the parser to find the open/close tags for head and body, and slice the html string using the indices.

this avoids allocating dom objects when reading and avoids the work of writing those dom objects back into strings afterwards

Testing

On AlignableWeb, used bombardier to make 100 request against /sitemap route.

Took memory and cpu profiles before and after, isolating extractDomElements function.

metric	before	after
cpu	300ms	170ms
memory	6 MB	1.8 MB

[reinabo]

Would be nice to unit test something like this, but we would require adding support for unit tests.

[dajinchu]

@reinabo added unite tests