configure.yml bootstrap-ids override leaves nodes with zero primary-network peers

Pre-existing, unrelated to the remote signer — surfaced during PR #25 e2e.

playbooks/l1/configure.yml sets global bootstrap-ids/bootstrap-ips to the L1 validator list. This replaces avalanchego's default Fuji beacon list (confirmed in the startup-logged config: bootstrappers = only our own validators). After phase 2, both nodes on the test stack had connectedPeers: 0 on the primary network and P: not connected to enough stake: 0% permanently — the health check fails forever, P-Chain stops following the network, and after any restart nodes can only re-peer with each other.

L1 consensus keeps working (self-contained), which is why this went unnoticed, but P-Chain-driven validator set changes, uptime tracking, and warp aggregation against fresh P-state are all degraded.

Fix direction: don't override the primary bootstrap list to point at L1 validators — L1 peer discovery goes through the P-Chain validator set; if explicit L1 bootstrap hints are needed they likely belong elsewhere (subnet config), not in global bootstrap-ids.

Read playbooks/l1/configure.yml and the startup-logged configuration from PR #25's e2e. Trace how global bootstrap-ids/bootstrap-ips are applied and how primary-network peer health is checked. Done means the test stack retains the default Fuji beacon list, nodes form primary-network peers, and the health check no longer remains at zero.