feat: implement deploy operator

[troian]

Rn manifest processing and deploy phases are tied together, which was a good solution at the time, now however it is causing quite a pain:

The problem

1. whenever processing manifest and deploying workloads fails for whatever reason, the provider tears deployment down including manifest. This is especially painful during the upgrade phase when a new version of the manifest is introduced.
2. The manifest controller does not support multiple manifest versions, which again creates huge pain during the upgrade process.
3. Hostnames and IPs processing are currently done in the deployment manager. Both features are cluster-specific, and yet again migration process with the current implementation is quite challenging.

The solution

Actual deployment of the Kubernetes objects must be done by the independent operator. It watches create/update/delete events of the manifest and makes changes to deployment based only on those events. The operator must support multiple manifest versions and update to the latest possible whenever configuration allows it. Failures during deployment must not affect the manifest.

The Kubernetes Client only processes manifests (derives hostnames and IPs if any) and stores/deletes them as CRD objects. Kubernetes Client must ensure the correctness of CRD to specification. The Client must support multiple manifest versions.

[brewsterdrinkwater]

December 19th, 2023:

• Artur will refactor some prior code
• Scott will take over once Artur has refactored

[brewsterdrinkwater]

January 9th, 2024:

• This is waiting for for feature discovery first.
• Akash Node discovery will come afterward.
• Code base is done. Just need to make sure it is stable.

[brewsterdrinkwater]

April 2nd:

• Changes are done, but some things need to be rebased.
• This will be pushed in the next few weeks. Ideally this should go out before provider configs.