Put babysitter configuration in ConfigMap.

[mwhittaker]

Previously, babysitter configuration was proto encoded, base64 encoded, and passed to the babysitter via environment variable. This PR instead puts the configuration in a ConfigMap:

apiVersion: v1
data:
  config.textpb: |
    namespace:  "default"
    deployment_id:  "e98a6946-49c2-4773-98c6-48b89a6c926f"
    trace_service_url:  "http://collatz-jaeger-60914c47:14268/api/traces"
    listeners:  {
      key:  "collatz"
      value:  20000
    }
  weaver.toml: |
    [serviceweaver]
    binary = "./collatz"
    rollout = "5m"
    colocate = [
      [
        "github.com/ServiceWeaver/weaver/examples/collatz/Even",
        "github.com/ServiceWeaver/weaver/examples/collatz/Odd"
      ]
    ]

    [kube]
    listeners.collatz = {public = true}
    repo = "docker.io/themwhittaker/"
kind: ConfigMap
metadata:
  creationTimestamp: null
  name: config-e98a6946
  namespace: default

The configuration includes the original weaver.toml file and a config.textpb file that includes a prototext encoded BabysitterConfig proto (which replaces ReplicaSetConfig). These files are mounted into the babysitter containers and passed to the babysitter as command line arguments:

containers:
- args:
  - babysitter
  - /weaver/weaver.toml
  - /weaver/config.textpb
  - github.com/ServiceWeaver/weaver/Main
  image: docker.io/themwhittaker/collatz:e98a6946
  name: serviceweaver
  volumeMounts:
  - mountPath: /weaver/weaver.toml
    name: config
    subPath: weaver.toml
  - mountPath: /weaver/config.textpb
    name: config
    subPath: config.textpb
volumes:
- configMap:
    name: config-e98a6946
  name: config

This change takes us a step closer to weaver kube exec, if that's something we end up doing. It also makes the generated YAML more readable and self-explanatory.

[rgrandl]

Looks much nicer. Thank you