feat(web): configure Prism role preferences and view capacity

[lukemaj]

Adds Settings → Prism (Model Router) directly below Providers. Each of the six roles shows easy, medium and hard lanes. Each lane stores an ordered list of enabled provider models and effort: primary first, then fallbacks. The same model can appear at different efforts. No capability tags.

The page uses the existing project/environment scope, preserves other lanes and role-kit fields, displays real provider usage windows and reset times, and links the recent-jobs placeholder to a thread's Agents panel. Saves and reset-to-environment share a guard that survives scope navigation and waits for the refreshed settings snapshot.

Closes #21; settings slice of toolboxmd/model-router#115. Rebased onto main after #22 merged. This PR changes no server or contract files.

Validation at deafe1b0bdbf2741c696c35f3a2b6acd06285dfc:

• 93 focused web tests; web typecheck; scoped lint and formatting; fork stack check.
• Independent GPT-6-Astra high review: no remaining blockers after correcting save concurrency across scope changes.
• Isolated live server (T3CODE_HOME under /tmp, separate ports), Playwright bundled Chromium: all 18 lanes, add/remove/reorder, distinct effort entries, independent drafts, save/reload, project override isolation and reset, real provider capacity, and Agents navigation. Zero page errors; narrow web viewport has no horizontal overflow.
• Screenshots captured locally with bundled Chromium. Image attachment to GitHub remains unavailable through the tools used here; no screenshot files were committed.

Upstream edits, recorded in both fork maintenance owners:

• apps/web/src/components/settings/SettingsSidebarNav.tsx
• apps/web/src/components/settings/settingsSearch.ts
• apps/web/src/routeTree.gen.ts (generated)

The route, page, logic, save state and tests are fork-owned files. No public website/SEO change. The fork has no version policy; upstream package versions remain unchanged.

Implemented with GPT-6-Astra in Codex; independently reviewed with GPT-6-Astra high in read-only Codex.

[github-actions]

Thread transfer impact

✅ Thread transfer remains within every enforced ceiling.

Provider	Metric	Main baseline	This PR	Impact	PR ceiling
Codex	Total thread wire	13.5 KiB	13.5 KiB	−28 B (−0.2%)	15.1 KiB	✅
Codex	Thread snapshot wire	7.1 KiB	7.1 KiB	−6 B (−0.1%)	7.3 KiB	✅
Codex	Live turn WebSocket wire	6.5 KiB	6.4 KiB	−22 B (−0.3%)	7.8 KiB	✅
Codex	Live turn WebSocket decoded	56.3 KiB	56.2 KiB	−44 B (−0.1%)	66.4 KiB	✅
Codex	Live turn messages	10	9	−1 (−10.0%)	21	✅
Claude	Total thread wire	13.5 KiB	13.5 KiB	0 B (0.0%)	15.1 KiB	✅
Claude	Thread snapshot wire	7.1 KiB	7.1 KiB	−7 B (−0.1%)	7.3 KiB	✅
Claude	Live turn WebSocket wire	6.4 KiB	6.4 KiB	+7 B (+0.1%)	7.8 KiB	✅
Claude	Live turn WebSocket decoded	57.0 KiB	57.0 KiB	0 B (0.0%)	66.4 KiB	✅
Claude	Live turn messages	9	9	0 (0.0%)	21	✅

Baseline: 6f9364f · PR result: deafe1b · Source CI: success

Scenario and decoded snapshot size

10 historical turns, 5 command tools per turn, 878.9 KiB retained MCP result per historical turn, and a 1.05 MiB retained result in the measured turn.

• Codex decoded thread snapshot: 113.9 KiB
• Claude decoded thread snapshot: 114.6 KiB

Updated in place by a trusted workflow. PR artifacts are strictly validated and never executed.