FindGlobalBMatrix should throw warning if no peaks indexed in a run

[RichardWaiteSTFC]

Describe the bug
FindGlobalBMatrix will not warn the user if peaks in one of the runs cannot be indexed - it just spits out a poor UB for that run - ideally if a run cannot be fitted the peaks would be excluded in the optimisation.

To Reproduce

# generate peak table
ws = LoadEmptyInstrument(InstrumentName='SXD', OutputWorkspace='ws')
axis = ws.getAxis(0)
axis.setUnit("TOF")  # need this to add peak to table

peaks = CreatePeaksWorkspace(InstrumentWorkspace='ws', NumberOfPeaks=0)
UB = np.diag([0.2, 0.25, 0.1])  # alatt = [3.8,4,10] 
SetUB(peaks, UB=UB)
SetGoniometer(Workspace='peaks', Axis0='-5,0,1,0,1')

peaks2 = CreatePeaksWorkspace(InstrumentWorkspace='ws', NumberOfPeaks=0)
UB = np.diag([0.3, 0.25, 0.1])  # alatt = [4.2,4,10] 
SetUB(peaks2, UB=UB)
SetGoniometer(Workspace='peaks', Axis0='5,0,1,0,1')

for h in range(0,3):
    for k in range(0,3):
        pk = peaks.createPeakHKL([h,k,4])
        peaks.addPeak(pk)
        pk = peaks2.createPeakHKL([h,k,4])
        peaks2.addPeak(pk)

peak_list = [peaks, peaks2]

FindGlobalBMatrix(PeakWorkspaces = peak_list,
    a=5, b=4,c=10,alpha=88,beta=88,gamma=89)
    
for ws in peak_list:
    nindexed, *_ = IndexPeaks(ws, Tolerance=0.15, CommonUBForAll=True, EnableLogging=False)
    print(f"{ws.name()} has {nindexed} indexed peaks")
    
# peaks has 9 indexed peaks
# peaks2 has 0 indexed peaks

Expected behavior
Would throw a warning and possibly not overwrite the UB on the workspace if it had one.

Screenshots

Platform/Version (please complete the following information):

• OS: [e.g. Windows, RHEL 7, Ubuntu, macOS]
• OS Version:
• Mantid Version [e.g. 6.0.0]

Additional context

[RichardWaiteSTFC]

After offline discussion, there are a few things I think that can be improved:

1. Single wrapper function around calling a child algorithm as per

   mantid/Framework/PythonInterface/plugins/algorithms/FindGoniometerFromUB.py

   Line 309 in bd91124

   def exec_child_alg(self, alg_name: str, **kwargs):

2. Change how initial indexing is found:
   • If no UB on any table then call FindUBUsingLatticeParameters on tables until successful then break
   • If there are UBs, copy UB to each workspace to get NUB x NUB triangular matrix of nindexed peaks - find best UB to use as a reference (leave exact method up to you - but the aim is to optimise the UB for as many runs as possible).
   • Once a reference UB has been found, index peaks in each table using this UB - if a table doesn't have enough indexed peaks then try and call FindUBUsingLatticeParameters and make_UB_consistent - if still nnot successful print warning and exclude table from optimisation.
3. After optimisation completed, loop over workspaces, call IndexPeaks and print warning for any runs which index < MIN_NUM_INDEXED (maybe also print at debug level how many peaks are indexed in each table)
4. Move try/except to be around the call to CalculateUMatrix in calcResiduals, so we can throw error telling user which workspace/run is problematic
5. Relax MIN_NUM_INDEXED = 2
6. In validateInputs require each table to have a minimum of 2 indexed peaks if a UB is present, or 6 peaks (don't need to be indexed) if no UB is present (atm just requires total of 6 peaks).

We don't need to address these in a single PR, let me know what you think/which bullet points you want to take on

[github-actions]

Closed by #38574.