chore: sync with upstream pingdotgg/t3code main

[yordis]

• Upstream now names every service after its application in the t3code namespace, so the fork drops its own service-name override and ledger entry 0023 in favor of upstream's names.
• Upstream's new OTEL environment reader covers only part of the fork's, so the fork keeps its reader to preserve exporter selection, batching, temporality, service version, the service-name warning, and per-value failure isolation.
• The relay deploy keeps the fork's runner and its skip when no credentials are configured, while adopting upstream's main-only guard and force input.
• The ACP prompt stall watchdog stays, rewritten in upstream's newer error-handling style so it does not diverge in shape.
• The background work banner keeps the fork's details popover and stop button alongside upstream's View agents entry.
• Upstream's new desktop and server OTEL tests are adjusted to the fork's behavior, where a bad header list drops only the headers instead of disabling export.
• Upstream's new Grok driver test provides the secret resolver the fork's drivers require.

---

^{Need help on this PR? Tag @codesmith-bot with what you need. Autofix is disabled.}

[cursor]

PR Summary

Medium Risk
Desktop changes touch startup, local IPC socket ownership, WSL env forwarding, and update install markers—user-visible on Linux and multi-instance setups; relay workflow changes affect production deploy paths.

Overview
Adds Linux .deb builds to release workflows, asset lists, README, marketing download page, and enables electron-updater for deb installs (alongside AppImage).

Desktop switches the packaged Electron entry to boot.cjs, which enables Node’s V8 compile cache before main.cjs and preloads it for the local backend in packaged builds only. Unix desktop app control sockets are reworked so multiple installs (e.g. nightly + preview) can share a state directory: newest app takes the path, close only removes its own inode, and reclaim re-binds when the socket file disappears (with directory watching).

OTEL renames the desktop service to t3code-desktop, adds service.namespace: t3code, tightens WSL forwarding of standard OTEL_* / T3CODE_OTLP_* vars, and expands tests for OTEL-vs-settings precedence.

Updates write a restart marker before install so the backend keeps managed tunnels across update restarts; marker is cleared on failed/interrupted installs.

Relay deploy gains workflow_dispatch with optional --force, main-only job guard, and RELAY_TUNNEL_CLEANUP_MODE from repo vars.

Android: subscription usage widget uses collection widgets (enabled on API 32+ only), scrollable rows, and refreshed copy; mobile UI scales toolbars, menus, FABs, and composer controls via useAndroidControlSizing.

Smaller follow-through: context menu checkbox items, t3code:// code cache for production scheme, Effect API cleanups (succeedNone, catchIf, etc.), and Linux arm64 release runner change to blacksmith.

^{Reviewed by Cursor Bugbot for commit a7473cf. Bugbot is set up for automated code reviews on this repo. Configure here.}

[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	+45 B (+0.3%)	15.1 KiB	✅
Codex	Thread snapshot wire	7.1 KiB	7.1 KiB	+14 B (+0.2%)	7.3 KiB	✅
Codex	Live turn WebSocket wire	6.4 KiB	6.5 KiB	+31 B (+0.5%)	7.8 KiB	✅
Codex	Live turn WebSocket decoded	56.2 KiB	56.3 KiB	+44 B (+0.1%)	66.4 KiB	✅
Codex	Live turn messages	9	10	+1 (+11.1%)	21	✅
Claude	Total thread wire	13.5 KiB	13.5 KiB	+12 B (+0.1%)	15.1 KiB	✅
Claude	Thread snapshot wire	7.1 KiB	7.1 KiB	+12 B (+0.2%)	7.3 KiB	✅
Claude	Live turn WebSocket wire	6.4 KiB	6.4 KiB	0 B (0.0%)	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: 897909f · PR result: a7473cf · 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: 114.0 KiB
• Claude decoded thread snapshot: 114.7 KiB

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