refactor(runtime): foundation for surface-context migration (#1237)

[gfargo]

What

Groundwork for the remaining open half of #1237 / #1136 — migrating workstation surfaces off threaded props onto LogInkRuntimeContext. The 0.72.1 state-cluster split is done; this starts the surface migration. No surface is migrated in this PR — it lands only the machinery the per-surface PRs consume, so it's behavior-preserving.

Changes

• Grow LogInkRuntimeContextValue with the set-once / app-wide values surfaces still receive as props: contextStatus + the injected render primitives h and components. (react/ink are dynamic-imported at boot, so a surface component can't import them — they ride in the context.) The provider value in app.ts populates the three new fields.
• useSurfaceRenderContext(React, panel) — rebuilds the SurfaceRenderContext a render*Surface fn already expects, from the context. The proven, snapshot-tested render logic moves unchanged; only how it sources inputs changes. panel selects main vs detail width.
• defineSurfaceComponent(React, renderFn, opts) — wraps a zero-extra render fn into a thin context-reading component (the common case for the first migration batch).

Tests

Fake-React harness (mirrors the established hook-test pattern, no renderer needed): read path, null-context throw, identity pass-through, main/detail width selection, and the component wrapper. header.test.ts updated for the grown value shape (the header is already a context consumer).

Validation

tsc --noEmit clean · full jest green (runtimeContext 9/9, header 6/6, 68 snapshots) · eslint 0 errors · rollup -c builds. (The 4 tsTreeSitterParser .wasm-backed failures are the known worktree WASM-grammar gotcha — unrelated, src/lib/parsers, fail identically on main here.)

Next

PR 2 migrates the 8 zero-extra surfaces (status, conflicts, remotes, submodules, reflog, pullRequest, issuesTriage, pullRequestTriage) onto these helpers. Full ladder in specs/surface-context-migration-plan.md.