Enhancement: Make --on-create hook timeout configurable (60s too short for yarn install)

[c-h-]

What happened

The --on-create hook has a hardcoded 60s timeout. For monorepos like UsableMachines/mono, yarn install in a fresh worktree can easily exceed 60s, causing the hook to time out.

Additionally, if the hook fails, the agent still launches into a potentially broken worktree (no node_modules, missing deps). This means the agent wastes its entire session trying to figure out why nothing works.

Current behavior

• --on-create runs in Phase 1, before agent launch (Phase 2) — ✅ correct sequencing
• Runs with cwd set to the worktree path — ✅ correct
• 60s timeout — ❌ too short for yarn install in large monorepos
• Hook failure doesn't abort launch — ❌ agent launches into broken state

Proposed fix

1. Make timeout configurable: --on-create-timeout <seconds> flag (default 60s for backward compat, but allow 300s+)
2. Add --on-create-must-succeed flag (or make it default): if the hook exits non-zero, abort the launch instead of continuing
3. Consider a config-level default in daemon config so users don't have to pass the flag every time

Use case

agentctl launch claude-code \
  --worktree ~/code/mono \
  --branch charlie/feature \
  --on-create "bash scripts/worktree-setup.sh" \
  --on-create-timeout 300 \
  -p "task"

Where scripts/worktree-setup.sh runs yarn install, env setup, etc.

Environment

• agentctl: latest (npm linked)
• Mono repo yarn install: 60-180s depending on cache state