Drop SqlJob and FlinkSqlJobReconciler

[ryannedolan]

Removed the SqlJob CRD, models, and related controller.

The idea behind the SqlJob CRD was to have non-functional metadata in Kubernetes for generic jobs, irrespective of an underlying implementation and runtime (e.g. Flink vs Spark). Then, a controller or external operator could translate SqlJobs into something concrete and deployable, e.g. FlinkDeployments, as was done by FlinkSqlJobReconciler.

However, this meant that every runtime (e.g. Flink vs Spark) would need to implement such a controller, even if they simply translate SqlJobs into something else. To avoid a lot of duplicate controller logic between runtimes, I'm dropping this intermediate CRD. Instead, the planner can generate concrete, deployable objects (e.g. FlinkDeployments) directly.

This simplifies several things without loss of power. I've opened #34 to track a newly relevant limitation: the planner can currently only generate one YAML object for each Resource, whereas a custom controller could create any number of downstream objects. But technically, an adapter could still implement a custom controller if needed, so we don't lose anything in the interim. And once #34 is fixed, we will be able to generate any number of downstream objects trivially -- without writing an entire controller.

Notice how simple the new hoptimator-flink-adapter is! It merely:

1. includes a SqlJob.yaml.template which generates FlinkDeployments.
2. registers the corresponding API endpoint (flinkdeployments).

I've added the flink adapter to the CLI s.t. !yaml will show FlinkDeployments instead of some generic SqlJob.

Testing

The !yaml command now generates FlinkDeployments instead of a non-functional placeholder SqlJob object:

0: hoptimator> !yaml SELECT * FROM RAWKAFKA."products"
...
---
apiVersion: flink.apache.org/v1beta1
kind: FlinkDeployment
metadata:
  namespace: {{namespace}}
  name: {{name}}-flink-job
...