CL: SIGTERM during startup while waiting for the EL forces a full redb repair on next start #466

App::start() opens the redb store before connecting to the EL. During the EL
retry window (up to 30 s) no SIGTERM handler is installed yet, so a
systemctl stop there kills the process with the store open and without the
quick-repair commit redb writes on Drop. The next start runs a full repair:
1h04 on a 103 GB testnet store. Seen on arc-node v0.8.0, testnet, follow mode.

Reproduce (EL stopped):

1. systemctl start arc-malachite, wait for the Failed to connect to Ethereum node retries.
2. systemctl stop arc-malachite during the retries.
3. Start the EL, then systemctl start arc-malachite: logs
   Database repair in progress: 0.00% instead of Database opened.

An orderly startup failure (execution engine unreachable ... Exiting...) does
not trigger it, since the store is dropped cleanly; only the unhandled signal
does. Opening the store after connect_and_resolve_chain_identity (first use is
State::builder) fixes it; verified on testnet. PR to follow.