Investigate SDP helm deployer replacement

As noted in the SDP review (https://confluence.skatelescope.org/display/SE/SDP+Review+for+reintegration+into+Skampi), the current implementation of the platform interface is a bit awkward - It is deeply coupled to a tool (Helm) that is:

• too complicated for what we want to do - which is generally just to deploy relatively simple "job" infrastructure (well, plus storage and possibly shared memory)
• doesn't cover everything we want to do - e.g. it is hard to easily get feedback about whether a certain deployment is healthy or not
• it requires Kubernetes to deploy infrastructure, which might potentially also not be the optimal choice for our use case
• is quite hard to maintain - many have complained that they do not like to write Helm charts in order to deploy workflows

Therefore it might be a good idea to investigate alternative implementations. The SDP architecture can support multiple deployment "backends" after all, so we can experiment.