Calculate docker instance label based on the hash of the config

[nikola-jokic]

Fixes #2647

The issue can occur when multiple runners are sharing the same docker socket, and they are executing within a container.
The well known directory is going to be the same for both runners, assuming they are executing the same image.

So when docker instance is calculated, the paths within the container are going to be exactly the same, one runner can remove resources owned by the other runner.

The fix uses the .runner file, which is unique per runner to calculate instance label, which should eliminate collision issues.

[fhammerl]

LGTM

This will make the docker instance label unique even if the runners sharing the same docker socket through docker-from-docker are under the same path in their respective containers.

Hashing the entire .runner file means that changes to this file (e.g. runnergroupname) will be reflected as a change in docker instance label as well - we found this acceptable as this change would happen outside of a running job, which means the docker network created based on this label in a previous job has been disposed.