Feedback regarding apb_adaptation

A couple of suggestions seeing https://gitlab.cern.ch/vcs/swarm-deployer-gitlab-runner/tree/apb_adaptation:

• RUNNER_TAG_LIST: this must stay an array. This means we will need to parse the APB parameters into an array, then re-encode the array as a string to pass it to the runner registration command. Even if the initial and final string are likely to be the same, we should still internally process the tag list as an array, because this is what this piece of data is. This will facilitate maintenance and future modifications (E.g. if we need the playbook to add tags to the user-supplied list obtained from APB parameters)
  NB: APB parameters have a type based on json-schame and it seems json schema supports arrays so you might be able to specify the APB parameter for the tag list as an array. Not sure what's the input format though, probably [ 'tag1', 'tag2' ]. Another option is type string with a textarea display type and one tag per line.

• normal builds and docker builds clusters could be 2 different "plans" for the APB

• CI: I understand we need to generate a complex label inside the Dockerfile. We do not want to have this label in our repo, this needs to be generated dynamically by the apb toolkit. Due to the constraints we have with docker builds on shared runners, we'll need to build in 2 steps:

  1. Prepare the Dockerfile (e.g. using apb prepare, or manually ), then save the result as build artifact
  2. build the docker image from the saved artifacts (rather than the repository source), using CONTEXT_DIR