By eliminating caching, and creating a large number of concurrent requests, I was able
to create similar behavior to that described in #798. The root cause seems to be reduced
network performance when a function is initially deployed.

This PR detects a GCF runtime, based on environment variables, and significantly extends
network timeouts.