Use full reconstruction on-the-fly for QEE tests

FYI @wbarter @ncooke @olupton @yqiu @cvazquez @hyin @dzuliani

Previously, the QEE tests were using the reco information that was stored in the file (from Brunel). This MR switches on the full, state-of-the-art reconstruction to run when we run our QEE lines. This allows us to benefit from the state-of-the-art performance and also to check that the state-of-the-art works for our use case.

Requires !22 (merged) for tests (and in general HltEfficiencyChecker) to be compatible with python3. A large min-bias run (>5000 events) may fail due to a tracking bug if you don't have the fix in Rec!2262 (merged), but the tests here on less events will pass by good fortune (the error is quite rare).

NB: the full reconstruction makes the lines ~3-4x slower, so I have reduced the number of events in our tests and increased the test timeout to account for this. The efficiencies and rates you get from the tests will be a bit less precise now, but we should always run HltEfficiencyChecker with lots of events anyway to quantify our physics performance, independent of making the tests pass.