spawn_dev.py: add an --effort flag, same shape as --model nf-18dh ← Beads

What's wrong

claude itself supports --effort <level> (confirmed: claude --help lists it). tools/office/spawn_dev.py passes --model through to the claude invocation it builds (launch_command, :972) but has no --effort flag at all — there's no argparse entry for it, and launch_command's argv never includes --effort even when a caller wants one.

Why this matters

Aedan asked for a spec round spawned at effort Max (2026-08-20). There was no way to satisfy that at spawn time — the office's /assign-family commands had to fall back to spawning normally, then SendMessage-ing the new session asking it to run /effort max on its own first turn. That works, but it's an extra round-trip and depends on the session actually doing it before starting real work, unlike --model which is set correctly from the first token.

Prescription

Add --effort <level> to spawn_dev.py's argparse (same shape as --model: optional, unvalidated — claude itself refuses a bad value, so no local allowlist to go stale), and thread it into launch_command's argv the same way model already is.

Test plan

spawn_dev.py <topic> --effort max --dry-run prints a claude invocation containing --effort max. Omitted, behavior is byte-identical to today (same discipline the model param's own docstring already states).