Change http_read_timeout and runtimeout defaults to be less than infinite

Puppet's http_read_timeout currently defaults to 0 which means infinite. If the agent sends an HTTP request and is waiting for a response, but the response never arrives (eg hard server crash, cable disconnect, router crash, etc), then the agent will hang indefinitely unless runtimeout is set to a non-zero default, or TCP keepalive detects if the connection is still valid.

In Puppet 6, we should change the http_read_timeout to something less. It also needs to be greater than the amount of time it takes the server to process catalog, report, etc requests.

We should also change the runtimeout to be something less than infinite as well.