Stop trying to measure relays not in the consenus

When we implemented #30727 (closed), we started to measure relays that are not in the last consensus. I think this is the cause why there's still around ~450 relays (see #40029 (closed)) failing to be measured because of circuit errors. We didn't realize that what we needed to do is to include these relays in the bandwidth file (and we already do, maybe we should just keep more days #40017 (closed)), but not in the list of relays to measure. The patch is simple, i'd try it.