Support HTTP keep-alive for health checks

[Castaglia]

The AWSHealthCheck functionality in mod_aws currently does not support HTTP keep-alive, and instead closes the connection after each request. Supporting HTTP keep-alive would allow for faster responses by avoiding the TCP connection overhead per request. This will be even more important once mod_aws supports SSL and/or HTTPS health checks; see Issue #14 .

[Castaglia]

With HTTP keep-alive comes long-lived TCP connections to the health check listener. Thus we'll need to have a maximum number of TCP connections at any one time, and timeouts for idle/unused connections (to prevent connection slot starvation). For example, if we haven't received/read a proper HTTP request from a connection within some amount of time (e.g. 60 seconds), close that connection.