Move over to controller-runtime over doing it ourselves

[davidcollom]

During testing we identified that there was significant memory utilisation when running version-checker for 2-3 days we saw that our test cluster was running at ~400+MB when it started off at around 40-60MB.

This PR changes over from manually implemented shared informers and managing the workqueue and its goroutines, to use controller-runtime, this means we don't have to manage as much code/logic and workers ourselves and is managed for us, allowing us to focus more on the purpose of version-checker.

We do believe that there is still a memory leak, however from testing (over 24 hour period) on the same cluster, that memory usage has increased at a much slower rate, additionally we've seen that goroutines are staying at a reasonable number, where as previously it was growing exponentially at, what was an alarming rate under some situations.

[davidcollom]

Some Screenshots of our findings.... Here's our last 2 days where version-checker was running for several days:

Here's a closer look at the memory utilisation:

And the GoRoutines:

After:

With a lot of this in mind, we've ensured that CPU usage/profile has not changed (%age is calculated towards 1 full CPU - I.E: 200m=0.2%) :

[davidcollom]

There's never a http.Client used within ACR - it creates its own - so removing.

[davidcollom]

I discovered a secondary leak where some metrics weren't cleaned up on pod deletion.