Slow and repeated network calls during ec2 functions

[hairyhenderson]

If I run a few ec2* functions on my laptop, usually it times out in ~5s and then subsequent calls are fast. But for some reason there are some environments where this doesn't happen, for example CircleCI, which does run in EC2, but they clearly restrict access to the EC2 APIs.

For example, running in a CircleCI debug session:

$  time (echo '{{ec2region}}{{ec2region}}{{ec2region}}' | ./gomplate)
unknownunknownunknown

real    0m15.011s
user    0m0.007s
sys 0m0.004s
$ time (echo '{{ec2tag "foo"}}' | ./gomplate)

real    0m10.279s
user    0m0.010s
sys 0m0.006s

I think the right approach is to do 2 things:

1. make an initial call to the EC2 Instance Metadata API, and track success - short-circuit further ec2 functions to use defaults after that (Short-circuit ec2 function defaults when not in AWS #60)
2. cache EC2 API calls to the same URLs so that we don't needlessly make repeated calls