This also fixes a potential problem when deduping SERPs against form history:
We're creating SERP URLs based on the lowercased version of SERP terms instead
of the terms as is. That means we won't detect example.com/?q=foo and
example.com/?q=FOO as being dupes, but I think we should.
I updated the test to check for leading/trailing spaces as well as case.
Differential Revision: https://phabricator.services.mozilla.com/D107164
51e0a45f0d