Refactor code for the move to Kubernetes

[nikodemas]

This PR together with dmwm/CMSKubernetes#1559 prepares the udp-collector for the move to the Kubernetes infrastructure.

FYI @vkuznet @leggerf @brij01

[vkuznet]

@nikodemas , I doubt that the changes you propose are valid. Let me break down for you the k8s requirements:

• the /health end-point should provide status of underlying service. In case of your changes you have UDP server serving UPD requests and separate goroutine (thread) of HTTP server using health end-point. If UDP server experience an issue the HTTP server knows nothing about it. Therefore proposed implementation does not capture the state of health of UDP server
• removing shell script has nothing to do with k8s, I suggest to keep them around for references. You may not need to pack them into a final image.
• you remove code with prometheus metrics which are useful to understand performance of the underlying service. In k8s you can rely on those to know what are the limits required to ru the service.

I suggest to re-factor the code in the following way:

• you may keep your HTTP server goroutine but it should do ping request to UDP server and receive back a reply. Doing this way it will probe the status of UDP server
• you may keep code under one main which will run both UDP and HTTP server but you must expose proper ports for them in k8s, one to serve UDP incoming requests, and another to expose HTTP to make health probes
• I suggest to put back prometheus metrics and use them within UDP server. What we should measure is performance of UDP server and not HTTP one. You may create a new UDP request for that, similar how original code is doing ping/pong exchange. A new UDP request will return UDP server prometheus metrics. And you HTTP server will need another end-point /metrics which will place UDP request to UDP server to get its metrics.