fix(desktop): extend remote cwd boot timeout

[konsisumer]

What changed and why

Per the reporter's clarification on July 9, 2026, the remote desktop path can still fail after Remote Hermes backend is ready with a renderer-side hermes:api timeout on current main. The desktop boot burst had already been widened for config/model/profile calls, but the remote desktopDefaultCwd() probe still used the Electron bridge's default 15s timeout. This PR moves that remote /api/fs/default-cwd request onto STARTUP_REQUEST_TIMEOUT_MS, so the post-connect boot path does not fail early on the leftover short timeout.

How to test

1. Run npm --workspace apps/desktop exec -- vitest run src/lib/desktop-fs.test.ts src/hermes.test.ts.
2. Launch Hermes Desktop against a remote gateway that is reachable but slow enough to exercise the post-connect startup burst.
3. Verify that after Remote Hermes backend is ready, the renderer can complete the remote default-CWD probe without tripping the old 15s hermes:api timeout.

What platforms tested on

• Not fully verified in this worker environment: the targeted Vitest command timed out locally without producing output.

Fixes #38061

[tonydwb]

Code Review Summary

Verdict: LGTM

What the PR Does

Extend remote cwd boot timeout to prevent startup failures on slow remote environments.

Assessment

• Small test-only change. Clean fix, well-scoped.

---

Reviewed by Hermes Agent