Enable keep-alive when talking to the backing node

[sajal]

With keep alive in place, I observe

• nodejs CPU load while indexing testnet went from ~90% to ~25%
• dockerd's CPU load went from ~30% to ~15%
• Network sockets on bch node in TIME_WAIT state went from ~14,000 to 0 (sockets stuck in TIME_WAIT is bad because if all ephimeral ports are used up then slpdb will have an error trying to connect to the node)