Fix port detection timeout for replaying nodes

When a node runs with replay_from, beekeepy waits for both HTTP and P2P
ports to become available. However, the P2P plugin only starts listening
after replay completes, which can take much longer than beekeepy's default
5-second initialization_timeout.

This caused intermittent FailedToDetectReservedPortsError failures in
mirrornet tests that replay large block_logs before syncing.

The fix updates settings.initialization_timeout to match the test timeout
when replaying, because beekeepy's _wait_for_app_to_start() reads the
timeout from settings directly rather than using the parameter passed to
_run().