#14493 Speed up deleting a summary ensemble

[magnesj]

Deleting a summary ensemble did work proportional to the number of realizations, in two independent places. Fixes #14493.

Measured

Deleting an ensemble from a project with many plots, in the running application. Before is 2322d9e97f on dev.

Measurement	Before	After	Change
deleteSummaryCaseCollection	2.1 s	0 ms	eliminated
~RimSummaryEnsemble	1.3 s	704 ms	1.85 x
Total	3.5 s	835 ms	4.2 x, 2.7 s saved

Before the change the two sections account for 3.4 s of the 3.5 s total, so nothing significant sits outside them. See #14493 (comment) for the full notes.

Delete curves for all cases in one pass

deleteSummaryCaseCollection() looped over every case, and for each case scanned every summary plot and called updateConnectedEditors() on every multi plot. Each plot was scanned once per realization, and the connected editors were updated once per realization and multi plot.

deleteCurvesAssosiatedWithCase() is replaced by deleteCurvesAssosiatedWithCases(), which takes a set of cases and returns whether anything was deleted. Each plot is scanned once, and the connected editors of a multi plot are updated only when that multi plot actually lost a curve.

This drops to 0 ms because nothing matches. The realization curves belong to RimEnsembleCurveSet::m_realizationCurves, not to the plot's RimSummaryCurveCollection::m_curves, so no curve is deleted and no editor is updated. Before the change, that same empty result cost 2.1 s.

RicCloseSummaryCaseFeature and RicCloseObservedDataFeature are updated to the new signature. RicCloseSummaryCaseFeature also scans each plot once instead of once per case.

Clear the referring curve sets instead of reloading them

~RimSummaryEnsemble() deletes the cases and then calls updateReferringCurveSets(), which runs RimEnsembleCurveSet::loadDataAndUpdate() on every referring curve set. Calling it after the cases are deleted is intentional, the curve sets are meant to end up empty, but reloading the curve data has nothing to read at that point. The reload also syncs UI fields, reloads the curve filters, rebuilds the address list, recomputes statistics over an empty case set and updates three legends, and it calls updateAll() on the parent plot once per curve set instead of once per plot.

clearReferringCurveSets() deletes the ensemble and statistics curves directly and updates each affected plot once.

The referring objects are queried as RimEnsembleCurveSet instead of walking all referring objects and casting. The previous loop discarded everything that was not a curve set, so the behaviour is unchanged.

curveSet->filterChanged.send() is kept, so RimAbstractCorrelationPlot, RimEnsembleWellLogCurveSet and RimEnsembleSurface still get the same notification as before. Sending a filter change while the ensemble is being deleted is questionable and it is the first thing to try if the remaining 704 ms is worth chasing, but removing it is a behaviour change and is left out of this fix.

Note

This touches ~RimSummaryEnsemble() and so conflicts with #14492, which replaces m_cases.deleteChildrenAsync() with m_cases.deleteChildren() in the same destructor. The resolution is to keep both changes, they are independent.

A dozen other types hold a PdmPtrField<RimSummaryEnsemble*>, among them RimAnalysisPlotDataEntry, RimParameterRftCrossPlot, RimRftTornadoPlot, RimOpmFlowJob and the ensemble histogram data sources. None of them are notified when the ensemble is deleted, before or after this change.