[PWGJE] jetCrossSectionEfficiency: cascade variants + hybrid-MC handling

- Rewrite the QC task as two ordering variants exposing the cumulative
  inefficiency through the analysis selection cascade:
  - processCrossSectionEfficiency (default): reco-collision-first
    (cascade = sel-bits via reco-coll EvSel bitmask, then NoSplit + zReco).
  - processCrossSectionEfficiencyBcBitsFirst: alternative ordering with
    sel-bits read from the truth BC (Preslice on JBCs) and reco-coll
    evaluated last; for systematic check.
- Cascade auto-adapts to {selTVX, selMC, selMCFull, sel8, sel8Full}; steps
  not in the preset pass through so bin count/order stays constant.
- Skips min-bias gap events via getSubGeneratorId() == mbGap when
  skipMBGapEvents = true (hybrid MB+JJ MC productions).
- Unified weighted/unweighted fills via mccollision.weight() (= 1 for MB).
- Outlier rejection via pTHat (stored ptHard, or weight-derived for legacy
  cases) gated by pTHatMaxMCP; named broken-sentinel constant.
- Restores all upstream-side Configurables (centrality*, trackOccupancy*,
  selectedJetsRadius, checkCentFT0M) for backward compatibility, even
  when unused on the pp MB MC nominal path.
- applyRCT toggles the RCT step independently; preset label is a
  Configurable.

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>

Author: Joonsuk Bae