Add paper reproduction catalogue and retained-data tools

[allamaprabhuani]

Summary

Add one paper-specific reproduction directory while retaining the existing
public examples, configuration schema and numerical solver behaviour.

• Index 38 manuscript and supplementary figure/table positions and 37 grouped
  numerical statements, including an explicitly pending recent profile analysis.
• Bind retained-data checks to the recorded manifest and document-edition hashes.
• Provide Q2, inverse and forward plotting adapters with separate saved-data and
  full-simulation status. Load historical plotting code from a trusted archive.
• Document the plotting environment, catalogue, output directories and remaining
  study requirements. Large result fields stay in the companion data archive.
• Add focused integrity, path-safety, adapter and catalogue regression tests.

Verification

• Full local public suite and new reproduction tests passed, 286 tests.
• Documentation built with sphinx-build -W -b html.
• The common CLI checked 5,875 retained files totalling 4,402,332,958 bytes.
• All three common CLI replot commands passed against a fresh archive extraction.
• Q2 checked 200 saved states. The inverse adapter regenerated six figures.
  The forward adapter regenerated standing-wave and subcycling figures and
  retained its numerical timing checks without adding an extra timing plot.
• Representative regenerated figures were visually inspected. Adapter checks
  compared pixels and rendered PDF content with the previously checked outputs.
• A Windows path-separator correction keeps logical input identities in POSIX
  form. Portable regressions cover Windows-style historical records and reject
  unsafe manifest paths. The inverse replot retains identical verified inputs,
  PNG pixels and rendered PDF content after the correction.

Scope

This contribution adds reproduction infrastructure and documented partial
coverage. The companion data record is being prepared. Remaining historical
inputs, environments, third-party reuse decisions and the complete simulation
routes are listed explicitly. No DOI, release tag or complete-rerun claim is
introduced.

The numerical Q2 integration is a separate contribution. This change leaves all
solver kernels, stopping criteria, defaults and existing example inputs intact.
Private manuscript sources, reviewer correspondence, research investigations,
raw cluster folders and unrelated teaching material are outside this diff.