Refactoring the interceptor

[arschles]

This PR fixes a bug in the interceptor in which it would not wait for a connection to be established if there was no server already waiting to accept conns, by adding a retry mechanism for the DialContext function in the HTTP proxy.

It does a few more things:

• Adding more testing to the proxy
• Refactoring the proxy to use standard net/http (gives it more control over the response lifecycle)
• Using the standard library test framework, because stretchr/testify/suite was not adding much (stretchr/testify/require is still in use though)
• Add a check in the proxy for the number of replicas in the deployment. If there are 0 replicas, it holds the request without trying to do any network operations until the replica count goes up
  • This is done in preparation for Use the external push gRPC protocol for external scaler => interceptor communication #97 - we want the interceptor to be able to use this knowledge to push a prompt to KEDA to start the scaling operation

Checklist

• Commits are signed with Developer Certificate of Origin (DCO)
• Have the interceptor watch the backing deployment and only send requests when they're at 1 or more replicas (when we get to Use the external push gRPC protocol for external scaler => interceptor communication #97, we can push to the scaler when we see a request come in and 0 replicas on the deployment)
• Fix tests
• Add more tests for the proxy handler
• Make the operator start up the interceptor with the proper new env vars
• Figure out the connection timeout tests (the ones that are failing right now)
• Make the interceptor time out on the waitFunc() (see newForwardingHandler)

Fixes #61

[arschles]

@tomkerkhove @khaosdoctor @zroubalik this was a bigger change and one of the big blockers for releasing alpha 2. Can you PTAL?

[khaosdoctor]

I think that's fine, there's a test failing though.

[zroubalik]

LGTM, there's failing test though

[arschles]

Ok, tests fixed. Thanks for the reviews @khaosdoctor @zroubalik !