Dynamically load Script Template and Code into Runner container

[nounder]

Creating container images for each template requires lots of maintenance and slow down iterations. Instead, we should use one container that fetches template files and script from Core on each startup. This will increase startup time which we can later on optimize by using H/2 for Runner-Core connection and running a lightweight service that caches all dependencies for every datacenter Runners are running.

[nounder]

Dependencies loading can be implemented directly in a Proxy service which we'll later on need. To further optimize costs Proxy and Dependencies services can be integrated into Core which has to run everywhere anyways.