Enable TCP keepalive on client socket

[TooAngel]

Problem

Chromecasts silently drop idle TCP connections — no FIN, no RST. The client's tls.connect socket doesn't have SO_KEEPALIVE enabled, so the OS never probes the peer, and a dead socket looks alive.

Symptom: after some idle period, the next send() writes into a socket that will never be read. Node reports success (the kernel buffer accepts the bytes), but nothing gets through. The application-layer heartbeat in node-castv2-client is supposed to detect this via missing PONGs, but does not fire reliably — see the long-standing thibauts/node-castv2-client#44. Callers end up hanging on their own timeouts.

Fix

One line: enable SO_KEEPALIVE on the TLS socket with a 10 s initial delay. When the socket is genuinely idle for that long, the OS starts sending keepalive probes; if the peer is gone, the OS closes the socket and Node emits close — which the existing onclose handler already handles.

this.socket.setKeepAlive(true, 10000);

Why it's safe

• Additive only. No existing code path is changed.
• During any active session, application traffic (heartbeat PING every 5 s, plus normal messages) keeps the socket busy — the OS never reaches the idle threshold, so keepalive probes are never sent.
• Only fires when the connection has been genuinely idle at the TCP level for ~10 s, which is precisely the failure mode we want to catch.
• Standard Node.js API; no new dependencies.

Verified against

Real Chromecast (Blaupunkt PMR100-style receiver). Before the change, first player.load after >5 min of idle hung for 8 s until the caller's timeout fired, then retried. After the change (patched at the consumer level as a preview), no hang and no retry needed.