Tidy up Statistics calculations with abstract StatisticsHelper class

[andrewbaxter439]

As a suggested refactoring, having a new abstract class inherited by all Statistics calculators could assist in collating repeated steps in a tidy way.

I've experimented with this in https://github.com/centreformicrosimulation/SimPaths/blob/feature/193-update-labour-supply-estimates-with-uc-flag/src/main/java/simpaths/data/statistics/StatisticsHelper.java but this may need review and formal restructuring to be included in main branches.

This feasibly allows calculation of routine sets of statistics in a tidy manner, e.g.:

       // mcs score
        CrossSection.Double personsMCS = new CrossSection.Double(model.getPersons(), Person.DoublesVariables.Dhe_mcs);
        personsMCS.setFilter(filter);

        // Calculate and set mean in one line
        setDhe_mcs_mean(calculateMean(personsMCS));

        // Calculate and set percentiles and medians in one line
        calculateAndSetPercentiles(personsMCS, this::setDhe_mcs_p_10, this::setDhe_mcs_p_25, this::setDhe_mcs_median, this::setDhe_mcs_p_75, this::setDhe_mcs_p_90);

[amorison]

The changes introduced via #503 already reduce the amount of churn associated with computing statistics, while increasing composability with e.g. caching mechanisms.

For instance, this example would become

var filteredPop = new FilteredCollection<>(model::getPersons, filter);
var mcsCs = new CrossSection<>(filteredPop, Person::getHealthMentalMcs);
var mcsStats = new Stats(mcsCs.get()).descrStats();

setDhe_mcs_mean(mcsStats.getMean())
setDhe_mcs_p_10(mcsStats.getPercentile(10.0));
// etc for other percentiles

Note that the longer term goal is to leverage the supplier API further and have something in this spirit:

class SomeClassHoldingStats {
    private Supplier<Stats> mcsStats;

    public SomeClassHoldingStats(SimulationEngine engine) {
        var filter = Filters.ageRange(25, 64);
        var filteredPop = new FilteredCollection<>(model::getPersons, filter)
            .oncePerSimTime(engine);
        var mcsCs = new CrossSection<>(filteredPop, Person::getHealthMentalMcs)
            .oncePerSimTime(engine);
        this.mcsStats = OnceUntil.timeChanges(engine, Stats.supplier(mcsCs));
    }

    public double mcs_mean() {
        return this.mcsStats.get().mean();
    }

    public double mcs_p_10() {
        return this.mcsStats.get().descrStats().getPercentile(10.0);
    }
    // etc for other percentiles
}

This leverages the Supplier API for cached deferred evaluation, and there is no longer a state to hold and update explicitly with the actual statistical values. The computations are naturally up-to-date, SomeClassHoldingStats only holds how to compute stuff and when it needs to be recomputed as internal state.

[andrewbaxter439]

This looks neat - will need to update the statistics calculations for the incoming UC parts to match this. Will do a branch and PR, will need some reviewing to get it working most likely!

[andrewbaxter439]

noticing that almost all the functionality is perfectly matched by #503 - looks good! Will close this PR and open up an additional one with some suggested additional things