fix-audit-407-r2: restore inventory + cross-refs for 19 :no-index: orphan classes

[igerber]

Summary

Holistic-audit fix-PR for #407 (docs: add :no-index: to page-level autoclass directives (PR 2 of 2)). Pilot worktree at /tmp/pilot-407-wt cherry-picked #407 onto its pre-merge base and ran iterative codex review (R0..R3 with gpt-5.4 + xhigh reasoning) until clean.

#407 added :no-index: to 78 page-level autoclass directives. The directive is correct for classes that also have a generated _autosummary stub (so the stub serves as the canonical inventory target). For 19 classes there was no autosummary entry anywhere in docs/api/index.rst, so :no-index: orphaned them — the class loses its only inventory registration, and any :class: cross-reference to it silently fails to resolve in the rendered HTML.

This PR adds the missing autosummary entries to docs/api/index.rst and commits the 19 auto-generated stubs (per repo convention; 111 stubs were already tracked).

Classes restored (organized by index.rst section):

• Panel Profiling (existing block): PanelProfile, OutcomeShape, TreatmentDoseShape, Alert
• Reporting (new block): BusinessReport, BusinessContext, DiagnosticReport, DiagnosticReportResults
• Boundary Local-Linear Estimators (new block): LocalLinearFit, BandwidthResult, BiasCorrectedFit
• Estimators + Results Classes (existing blocks): StaggeredTripleDifference, StaggeredTripleDiffResults, TWFEWeightsResult
• HAD Pretest Workflow (new block): HADPretestReport, QUGTestResults, StuteTestResults, YatchewTestResults, StuteJointResult

Cross-refs explicitly repaired:

• docs/api/index.rst:108 → :class:~diff_diff.PanelProfile``
• docs/api/profile.rst:7,21 → :class:Alert/:class:OutcomeShape / :class:TreatmentDoseShape``
• docs/choosing_estimator.rst:19 → :class:~diff_diff.StaggeredTripleDifference``
• docs/api/had.rst:163 → :class:~diff_diff.HADPretestReport``

Build verification:

Verified locally via make -C docs clean && make -C docs html SPHINXOPTS="-W":

• Build succeeds (0 warnings, -W enforced)
• 19 new _autosummary stubs generated using the custom class.rst template added in docs: custom autosummary template + residual triage (PR 2.5) #410 (auto-emits :no-members:)
• Cross-refs in the rendered HTML resolve to _autosummary/diff_diff.<Class>.html

No methodology impact — pure inventory/navigation restoration.

Test plan

• CI green

[github-actions]

Overall Assessment

✅ Looks good — no unmitigated P0/P1 findings. The inventory/autosummary repair appears sound; the only issue I found is a P2 documentation regression around methodology-link navigation and TODO tracking.

Executive Summary

• Cross-checked the touched methodology prose against docs/methodology/REGISTRY.md:L474-L666, docs/methodology/REGISTRY.md:L1424-L1694, and docs/methodology/REPORTING.md:L56-L143; no undocumented estimator or inference deviation was introduced.
• The new autosummary coverage in docs/api/index.rst:L20-L249 is internally consistent, and I found no remaining top-level :no-index: orphan classes after this change.
• Severity: P2 The PR replaces live methodology references with plain inline file paths in the reporting docs/tutorials, even though the Markdown methodology files still are not rendered as Sphinx pages.
• The same PR also removes the Testing/Docs TODO that had been tracking that unresolved rendering/navigation gap, so the debt becomes untracked rather than fixed.
• No code quality, performance, maintainability, or security issues beyond that docs item.
• Review note: this was a static diff review; I did not run Sphinx/tests in this environment.

Methodology

• No findings. The touched reporting-layer, SyntheticDiD, and dCDH claims remain consistent with docs/methodology/REPORTING.md:L56-L143, docs/methodology/REGISTRY.md:L474-L666, and docs/methodology/REGISTRY.md:L1424-L1694.

Code Quality

• No findings.

Performance

• No findings.

Maintainability

• No findings beyond the documentation-link regression below.

Tech Debt

• No separate findings beyond the removed tracking item below.

Security

• No findings.

Documentation/Tests

• Severity: P2 Impact: docs/api/business_report.rst:L48-L66 and docs/api/diagnostic_report.rst:L14-L21 now render docs/methodology/REPORTING.md as inline code, not a hyperlink. But docs/conf.py:L23-L33 still has no Markdown-page support, so REPORTING.md/REGISTRY.md are not first-class Sphinx pages. The changed tutorials use the same non-link pattern for REGISTRY.md (docs/tutorials/18_geo_experiments.ipynb:L43, docs/tutorials/19_dcdh_marketing_pulse.ipynb:L17, docs/tutorials/19_dcdh_marketing_pulse.ipynb:L41). At the same time, the Testing/Docs debt list no longer tracks the follow-up (TODO.md:L137-L143). Concrete fix: either restore the prior external links until these methodology docs are actually rendered, or implement the tracked fix properly (enable Markdown pages in Sphinx or convert them to .rst, then use :doc:). If rendering stays deferred, restore the TODO entry instead of deleting it.