[PWGJE] jetCrossSectionEfficiency: rewrite with CollRecoFirst + BcBitsFirst cascade variants

Replaces the upstream processCrossSectionEfficiency[Weighted] (jet-pT QC over
expanded sel8 bits) with two ordering variants that expose the cumulative
inefficiency through the analysis selection cascade:

- CollRecoFirst: reco collision required first, BC bits read from the reco-coll
  EvSel bitmask, NoSplit + zReco at the end.
- BcBitsFirst:   BC bits read from the MC truth BC, truth-side SBP via Preslice,
  reco collision evaluated last (alternative ordering for systematic check).

The cascade adapts to one of {selTVX, selMC, selMCFull, sel8, sel8Full} via the
eventSelections preset; the kNoITSROFB and kNoSBP steps are skipped (pass-through)
when the preset does not include them so the bin order/count stays constant.
applyRCT toggles the RCT step independently.

Outlier rejection uses pTHat (stored mccollision.ptHard() or weight-derived
via simPtRef/pTHatExponent) gated by pTHatMaxMCP, mirroring the convention in
jetSpectraCharged. Broken-sentinel value is named (kBrokenPtHardSentinel)
across both files. RCT preset label is now a Configurable.

Dropped configurables vs the upstream task (analysis scope is pp MB MC):
centrality*, checkCentFT0M, skipMBGapEvents, trackOccupancy*, selectedJetsRadius.

Validated on LHC23k4i pp 13.6 TeV MB MC (cascade smoke) and against the LHC25a2
JJ MC anchor (weighted-bind fix in jetSpectraCharged.cxx; negative control on
unpatched daily-20260420 reproduces the original FATAL).

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

Author: 2 people