feat(rate-of-closure): link launch monitor scatter selection

[d-sorgcodexagent]

Summary

Advances #4433 from initial visibility into matched direct visualization interaction for the React and PyQt Launch Monitor Analytics tabs.

• preserves each retained row by dataset-generation plus zero-based ordinal; no player identity is inferred from shot/session/vendor fields
• links both scatter views through pointer and keyboard selection with Home/End/arrows/Escape, selected-marker emphasis, and accessible retained-row status
• strictly aligns portable decimal/CSV/JSON projection semantics across Python and TypeScript and rejects ragged, nested, nonfinite, radix, Unicode-digit, and unsafe values
• keeps raw retained data available to the analysis missing-data policy while deterministically capping display geometry at 2,000 marks and failing closed above 250,000 retained rows
• preserves selected rows through display decimation, distinguishes missing versus decimated availability, and computes nearest pointer selection in rendered pixels
• ensures selection/reset do not rerun statistics; predictor/outcome/request edits invalidate stale analysis results and exports until a fresh Run
• adds responsive rendered interaction and selected-state PyQt diagnostics without promoting artifacts to approved goldens

Stack

• Base PR: feat(rate-of-closure): keep primary visualizations above the fold #4435, exact branch head 4dde158109d85df2bf876d73f0c384d482fb989d
• Feature commit: 871eb74869f4f40bc083ea648871c014e933bbcf
• Final static correction: 4d80c66a230437dd27c4699591cb42ed97d86b50
• Normal linear ancestry; no rebase, retarget, or rewrite.

Evidence

• Rate Python: 1,266 passed
• React Vitest: 906 passed
• focused Python/TypeScript: 20 / 18 passed
• production Playwright linked interaction + visibility: 3 passed across 1440x900, 1280x720, and 390x844
• PyQt full 18-tab audit at 100% and 150% DPI: passed, including selected-row diagnostic
• 100,000-row planner: about 86–90 ms, exactly 2,000 displayed marks
• Vite build, TypeScript, ESLint, Ruff, format, diff, and changed-file <=400 LOC: green
• independent exact-head review: GO

Honest boundary

This remains a presentation-only linked view. Dataset fingerprints are runtime-local and nonportable; raw authority/export is not decimated. Approved visual goldens, formal assistive-technology review, archive integration, generalized shared interaction roles across the other tabs, all-state geometry, user approval, downstream validation, and protected publication remain open. #4433 is not complete.

[d-sorgcodexagent]

Reviewed correction head

PR #4440 now points to exact 7030a846f84ace0dde1ee0df0bf191ed7b3b3ed3.

Late adversarial review closed before publication:

• overflow-safe normalized plotting for ±MAX, constant maxima, near-ULP maxima, signed zero/subnormals, and zero paired points in both renderers;
• atomic dataset-bound defaults/results/exports with browser import epochs and PyQt generation-tagged callbacks;
• strict 8 MiB / 250k-row / 256-column / 2M-dense-cell / 64 KiB-per-field UTF-8 import boundaries, duplicate-key rejection, fatal UTF-8, malformed CSV normalization, suffix-before-read, and no global CSV field-limit mutation;
• identical well-formed-Unicode scalar rules for JSON keys/string values, including lone-surrogate rejection, valid supplementary pairs, and normalized duplicate detection;
• exact hosted MyPy boundaries fixed.

Broad prior evidence remains: 1,271 Rate tests, 910 React tests, production React/PyQt rendered audits, build/static gates, and independent exact-head GO reviews. Fresh protected checks are queued/running normally; no merge-completion claim is made and #4433 remains open.