chore: merge upstream pingdotgg/t3code through d7c711503f

[andrewcai8]

Merges 60 upstream commits. Notable ones: a /compact stopped by the Claude usage limit now fails instead of hanging (pingdotgg#19583), the mobile usage chart legend wraps instead of pushing the end date off-screen (pingdotgg#18203), and colored terminal output no longer copies history one character at a time (pingdotgg#21209).

This stops at d7c711503f on purpose. It is upstream's newest commit with green CI. Every upstream CI run after it fails, about 200 in a row through e2cc34f1e4, including typecheck: GrokAdapterV2.test.ts imports Path twice, and two web tests import react-test-renderer, which apps/web doesn't declare. Those commits come in with the next sync once upstream is green again.

Conflicts:

• runtimeLayer.test.ts: kept the fork's held-queue test and added upstream's held-queue continuation-ordering test.

Two fork tests needed updates for upstream behavior changes:

• RepositoryIdentityResolver.cloud.test.ts: upstream now keeps a repository's display casing (fix(server): preserve repository display casing pingdotgg/t3code#19202), so owner is T3Tools. The host still lowercases every repository through canonicalRepository() before matching its config, so cloud repository lookup is unaffected.
• ProviderAccountSwitch.test.ts: upstream renumbers a queued run so it lands after a continuation that ran first (fix(server): queued messages land after a resume that ran before them pingdotgg/t3code#17773). Execution order is the same as before, and the queued run still runs on the new login. Only the ordinals moved from 2/3 to 3/4.

Verification: typecheck is clean for server, web, mobile, client-runtime, contracts and shared. Server tests for the files both sides changed pass: runtimeLayer, ClaudeAdapterV2, ThreadLaunchService, AgentSessionScanner, OrchestratorMcpService, Orchestrator (546 tests). One AgentSessionScanner symlink test fails only with macOS's default /var temp dir and passes with a real-path TMPDIR.

🤖 Generated with Claude Code

[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	5.0 KiB	5.0 KiB	0 B (0.0%)	6.8 KiB	✅
Codex	Thread snapshot wire	3.8 KiB	3.8 KiB	0 B (0.0%)	4.9 KiB	✅
Codex	Live turn WebSocket wire	1.2 KiB	1.2 KiB	0 B (0.0%)	2.0 KiB	✅
Codex	Live turn WebSocket decoded	20.9 KiB	20.9 KiB	0 B (0.0%)	29.3 KiB	✅
Codex	Live turn messages	2	2	0 (0.0%)	8	✅
Claude	Total thread wire	5.0 KiB	5.0 KiB	−31 B (−0.6%)	6.8 KiB	✅
Claude	Thread snapshot wire	3.8 KiB	3.8 KiB	0 B (0.0%)	4.9 KiB	✅
Claude	Live turn WebSocket wire	1.2 KiB	1.2 KiB	−31 B (−2.5%)	2.0 KiB	✅
Claude	Live turn WebSocket decoded	21.2 KiB	21.2 KiB	−41 B (−0.2%)	29.3 KiB	✅
Claude	Live turn messages	2	1	−1 (−50.0%)	8	✅

Baseline: b70f25c · PR result: 786ec06 · 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: 108.5 KiB
• Claude decoded thread snapshot: 108.8 KiB

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