fix(cli): add timeout to ag run session creation fetch (#3247)

[OneStepAt4time]

Summary

Fixes #3247 — ag run hangs silently when session creation is slow.

Problem

ag run has no timeout on the POST /v1/sessions fetch call. When the server is slow (ACP prompt delivery with BYO-LLM proxies), the client hangs indefinitely with no feedback — exactly what broski hit as our first external tester.

Root Cause

Two compounding issues:

1. No client timeout: The fetch call in handleRun() has no AbortSignal.timeout() — it hangs forever if the server is slow
2. Server-side: v0.6.6 has 15s ACP timeout (fixed on develop at 120s), but even with 120s the client should have its own timeout

Fix

• Added AbortSignal.timeout(120_000) to the session creation fetch call
• Added specific error message for timeout: tells user to try increasing AEGIS_ACP_PROMPT_TIMEOUT_MS
• If timeout fires, user gets a clear error instead of silent hang

Important note

Broski installed v0.6.6 which does NOT include the ag run command (added after v0.6.6 in #3043). This fix needs to ship in v0.6.7 release along with the 120s server-side timeout (#3243).

Verification

tsc --noEmit  ✓
npm run build  ✓
npm test       ✓ (5 pre-existing failures unrelated to this change)

Commit: 7dadd32

[aegis-gh-agent]

Approved — all CI green.

Clean fix: AbortSignal.timeout(120_000) on session creation fetch + AbortError handling with actionable error message.

Root cause analysis is solid — broski hit v0.6.6 which predates ag run entirely. This fix needs v0.6.7 release along with the 120s server-side timeout (#3244).

P1 release blocker resolved client-side. Shipping v0.6.7 is the next critical step.