fix(gateway): verify process is alive after restart before reporting success

[sridhar-3009]

Summary

Closes #301

ohmo gateway restart called start_gateway_process() and immediately
printed "restarted (pid=N)" — no verification that the subprocess was
still alive. If the gateway crashed on startup (bad config, port
conflict, missing dependency), the command silently reported success
while the service stayed down, requiring a manual container restart.

Root cause in start_gateway_process:

• stdout/stderr are redirected to gateway.log so no output
  appears in the calling terminal.
• subprocess.Popen returns a handle instantly; the PID is returned
  immediately without checking whether the process survived.

Fix: After spawning, poll _pid_is_running(process.pid) for up to
~1.5 s in increasing intervals (0.1 s → 0.2 s → 0.4 s → 0.8 s).
This gives the process enough time to initialise before declaring it
healthy, while still catching immediate crashes. If no poll succeeds,
raise RuntimeError with the PID and log file path.

Update gateway_restart_cmd and _maybe_restart_gateway to catch that
error and surface a clear message (non-zero exit for the CLI command).

Test plan

• Start gateway normally — verify ohmo gateway restart still prints restarted (pid=N) and the service is up.
• Simulate a startup failure (e.g. invalid gateway config) — verify restart now prints an error pointing to the log file instead of falsely reporting success.
• Verify gateway start (not modified) still works unchanged.