Report per phase measurements for correct phase

[sthelen-enqs]

If/when a device registers PhaseSpecific measurements only for one specific phase (or registers them for all phases but only provides measurements for one of them), MeasurementPhaseSpecificDataForFilter() returns a []float64 containing only 1 value with no way for the API user to know for which phase the measurement was meant. This is an issue in e.g. ma/mpc's PowerPerPhase which would then return an array with less than 3 members and callers would then not know to which phase the measurement corresponds.

This PR changes the return type of MeasurementPhaseSpecificDataForFilter to always contain one entry per phase so that for a device which only reports data on phase B, it would return []float64{0, 10, 0} instead of []float64{10}. The advantage of this approach is that it doesn't change the return type of MeasurementPhaseSpecificDataForFilter or any of its callers, the downside is that callers can't (easily) distinguish devices that are reporting 0 for a phase and devices that aren't reporting a phase at all. The alternative would be to change the return type from []float64 to map[ElectricalConnectionPhaseNameType]float64 and then going through all the callers to make them use/return that as well.

I'm not particular to either approach so I picked the one that changed the surface API less.

[sthelen-enqs]

After a discussion, we'll rewrite this to return the map[ElectricalConnectionPhaseNameType]float64 to be more explicit about whether phases were written to or not.

[andig]

Another and easier to use option might be a [3]*float.

[sthelen-enqs]

I've updated this PR to return a map[ElectricalConnectionPhaseNameType]float64.
There is one incidental change which may be of interest. The acPowerTotal measurements must now have an acMeasuredPhases electricalConnectionParameterDescriptionData set for the MA MPC/MGCP usecases to pick up/report the value via the Power() functions. These are required values in the MPC/MGCP specification and should therefore not affect communication with real devices (in theory), but I had to adjust two test cases to add the required parameter.