Benchmarks with other parsers

[sylvinus]

We should make sure we are not leaving some performance on the table compared to other solutions!

[sylvinus]

These first naive tests are just counting divs and gumbocy is (purposely) ignoring a lot of the html, so the results are to be taken with a grain of salt. Still, lxml is 2x faster (but not HTML5-compilant), and all the other Python alternatives range from 7x to 20x slower.

benchmark_gumbocy: 0.555 seconds
benchmark_gumbo_bs3: 8.960 seconds
benchmark_lxml_raw: 0.254 seconds
benchmark_html5lib_bs4: 9.977 seconds
benchmark_htmlparser_bs4: 3.534 seconds