content_script.js blocks the main thread of the content process scanning the entire document for elements that, for most pages, aren't there

[mikeconley]

Here are the CSS selectors for the things that the add-on scans for.

A number of these queries are site-specific - however, it looks like we scan every page for them. I suspect we can skip a lot of this scanning by checking the domain that we're on, and culling rules that don't apply. I believe that would eliminate all but 2 of the LOGIN_PATTERN_DETECTION_SELECTORS selectors, and all but 1 of the SHARE_PATTERN_DETECTION_SELECTORS selectors.