feat(virtual-table): logs/streaming-specialized VirtualTable (F2-A1 design-study cell)

[gethinwebster]

Summary

Design-study cell F2-A1 of the VirtualTable exploration: a config-driven, logs/streaming-specialized additive component that coexists with Table (not a rewrite). One of three parallel winner threads from the study (see also #4738 F1-A1 generic config-driven, #4739 F1-A2 generic compound).

What's here

• view/viewConfig discriminated union: standard / patterns / raw surfaces (single source of truth).
• Built-in live tail (controlled follow, pin-to-newest, release-on-scroll-away, rate-limited polite append, reduced-motion-safe).
• Two-mode filter (subset / mark-in-place) with a materialised, counted match-indicator column and non-visual match conveyance (WCAG 1.4.1) + keyboard next/prev-match.
• R-EXPAND: arbitrary non-tabular expanded content via a render slot with log-record(~300px) / pattern-detail(~150px) presets layered on the slot (not a closed set); valid grid-child expanded region, measured variable heights.
• Full-dataset windowing (single owned scroll container, opt-in measure-on-first-entry + offset correction), full-dataset aria-rowcount/aria-rowindex (filtered length in subset mode), materialised disclosure column.
• CloudWatch Logs Insights standard + patterns + raw demo pages; unit + a11y (axe/keyboard/SR) tests; usage docs.

Testing

• npm run build green; npm run lint green (0 errors); npm test (virtual-table) — 6 suites, 71 tests green (incl. axe across all views, keyboard, live-append debounce, indicator-column colindex coherence).

Human-gated (NOT in scope)

• Official Cloudscape API review and design/release sign-off.
• Component naming, and convergence across the three design-study threads.
• Follow-ups: prefer the internal LiveRegion component over the raw aria-live node (one lint warning); full-run coverage thresholds.

Draft, one cell of a design study — opened for discussion, not merge.