Create sector-aware RHIME outputs and PARIS-compatible total outputs

[brendan-m-murphy]

Goal

Convert modern multi-sector InferenceData results into outputs that preserve sectors while still
producing production PARIS outputs.

Tasks

• Reconstruct sector-specific posterior flux traces or summaries outside the PyMC graph.
• Add total posterior flux by summing sector posterior flux when all sectors share the same basis.
• Preserve sector-specific modelled-observation contributions where practical.
• Route total outputs through PARIS output creation.
• If PARIS cannot carry a sector coordinate, emit a PARIS-compliant total output plus a separate
  sector diagnostic NetCDF.

Acceptance criteria

• Output contains sector posterior flux and total posterior flux.
• Total flux equals the sum of sectors for a synthetic fixture.
• PARIS flux and concentration outputs work for total outputs.
• Sector diagnostics are saved or returned in a documented format.

[brendan-m-murphy]

PR #420 adds minimal multi-sector diagnostics, but does not complete this issue.

Important detail from real-data testing: sectors may share basis regions, but they do not share flux fields. So postprocessing should not create a fake single total scale factor by summing/stacking x_<sector> blindly. The correct path is to reconstruct each sector posterior flux from its own x_<sector>, sector prior flux, and shared basis, then sum posterior fluxes to produce total outputs.

Remaining work:

• sector posterior flux traces/summaries
• total posterior flux from summed sector fluxes
• total PARIS/basic outputs
• documented sector diagnostics format