Check workflow performance

[jl-wynen]

Need to ensure that we can process the expected data volume with the available memory and fast enough for Beamline.

[SimonHeybrock]

See also scipp/essreduce#12. Note the options for growing monitors and detectors independently, as they may result in independent performance bottlenecks. -> See #45, then we can proceed here.

[SimonHeybrock]

Baseline for Geant4 workflow:

2.47 seconds load_geant4_csv
4.75 seconds load_geant4_csv
3.00 seconds extract_geant4_detector_data
0.62 seconds normalize_by_proton_charge
6.11 seconds extract_geant4_detector_data
4.38 seconds add_scattering_coordinates_from_positions
0.35 seconds apply_masks
0.99 seconds normalize_by_proton_charge
4.36 seconds convert_to_dspacing
5.20 seconds add_scattering_coordinates_from_positions
0.23 seconds apply_masks
2.94 seconds convert_to_dspacing
9.75 seconds focus_data_dspacing
10.12 seconds focus_data_dspacing
1.19 seconds normalize_by_vanadium_dspacing

I artificially increased the data size by modifying the data after loading:

def extract_geant4_detector_data(
    detector: RawDetector[RunType],
) -> RawDetectorData[RunType]:
    """Extract the histogram or event data from a loaded GEANT4 detector."""
    data = extract_detector_data(detector)
    data = data.bins.concatenate(data)
    data = data.bins.concatenate(data)
    data = data.bins.concatenate(data)
    data = data.bins.concatenate(data)
    data = data.bins.concatenate(data)
    data = data.bins.concatenate(data) 
    data = data.bins.concatenate(data)  # ~ 10 GByte (sample run)
    return RawDetectorData[RunType](data)