get_match_for_measurement should follow precedence rules of matching qubit

[MarquessV]

Per this discussion the precedence for determining the best match for a Measurement to a vector of MeasureCalibrationDefinition is as follows:

1. Match fixed qubit
2. Match variable qubit
3. Match no qubit (I don't recall this case but it's supported here)

The tiebreaker among calibrations of equal precedence is "last wins."