Fix goroutine leak

[piotrmiskiewicz]

The number of goroutines is constantly growing. It looks like the AvS client (http with oauth) is created for every http call. It should be changed to use shared one. The initial AvS implementation was done this way, but there was a problem with token refreshing (see https://github.com/kyma-project/control-plane/pull/268/files).
AC:

• no goroutines leak

In case of extracting AvShttp client to shared one - test the implementation for more than 24 h to check if it works fine (token refreshing etc.)

[MarekMichali]

One way to test it on dev without promoting it to stage is to add the pr-deploy label to the promotion PR. This will allow you to push the tested KEB version to dev with promotion to stage blocked, and will also make the metrics available on the dashboard so we can see how it behaves. A good idea would be to do this on Friday and revert on Monday, so we get over 48 hours of testing without interfering with the normal KEB releases that happen during the week.

We should also create a custom KEB release from a fork for this testing scenario, by building and hosting the KEB chart there as we do upstream, without interfering with the upstream KEB release process. After successful testing, everything should be merged form the fork to upstream and released there.

[MarekMichali]

The leak is somewhere in AVS_Create_Internal_Eval_Step during both provisioning and deprovisioning. This was confirmed by disabling this step. After disabling it, the goroutines count stopped growing. This issue was also only seen on dev.

[MarekMichali]

Creating AVS external evaluation is disabled on all environments, see #139

[MarekMichali]

We decide to leave this issue for now, as it only occurs on dev.