Update UB matrix if IndexPeaks called with CommonUBForAll=False and table contains only peaks from one run

[RichardWaiteSTFC]

Is your feature request related to a problem? Please describe.
If a peak table contains only peaks from a single run then IndexPeaks called with CommonUBForAll=False will optimise the UB but this won't be stored on the workspace

mantid/Framework/Crystal/src/IndexPeaks.cpp

Lines 503 to 506 in 56f61ef

	if (peaksPerRun.size() < 2) {
	g_log.warning("Peaks from only one run exist but CommonUBForAll=True so peaks will be indexed with an optimised "
	"UB which will not be saved in the workspace.");
	}

Describe the solution you'd like

Update UB matrix if IndexPeaks called with CommonUBForAll=False and table contains only peaks from one run - in this way if RoundHKLs=False
peak.getQSampleFrame() = 2*np.pi*peaks.sample().getOrientedLattice().getUB() @ peak.getHKL()

Perhaps for backwards compatibility would require an additional parameter e.g. UpdateUB = False - but it's difficult as this only applies to some peak tables (with peaks from 1 run).

Describe alternatives you've considered
Can call CalculateUMatrix with indexed peaks and observed lattice parameters, this does seem to produce better results sometimes than e.g. FindUBUsingLatticeParameters. OptimizeCrystalPlacement also often seems to find goniometer offsets that improve the accuracy of the UB.

Additional context
How to treat tables with peaks from multiple runs? Could update the goniometer matrix for peaks from the same run such that
R_old * UB_new = R_new * UB_old?
Not sure this is intuitive/desired behaviour, though it might help with calibrating goniometer angles, axes and offsets

[RichardWaiteSTFC]

@zjmorgan what do you think of this? We don't have to do it, but it seems a shame to throw away a better UB!

[zjmorgan]

Seems like a good idea