I converted the cps2 tests to use async/await, since it was easier
for me to debug the breakages introduced by my changes that way.
There are a few differences hidden in there, though - most notable
is that I also modified the direct db access in the tests to use
the Sqlite.jsm wrapper. This was to avoid reaching into the
connections internals in order to get the underlying connection.
MozReview-Commit-ID: DwJ5hfyvAOf
6c2133cb4e