Track "real" pfsConfigId when taking odd and even quartz flats.

The 2d pipeline expects to match science exposures to their flat and arc calibration exposures using their common pfsConfigId.

But when taking odd and even fiber quartz exposures, half the cobras will be behind the spots so the actual configId will be wrong.

Either lie about the configId, or, almost certainly better, include a card with the configId for the full field. Oh, `W_PFIDFL` or something equally ncmprhnsbl.