appliance: fix panic on reconcile

[craigfurman]

It needs an event recorder.

Test plan

Manually tested by starting the appliance with go run ./cmd/appliance, and uploading an annotated configmap to a locally-running kubernetes cluster, to start the reconcile loop.

[craigfurman]

Example of how this shows up in events:

❯ kubectl -n test describe configmaps sg

Name:         sg
Namespace:    test
Labels:       deploy=sourcegraph
Annotations:  appliance.sourcegraph.com/currentVersion: 5.3.9104
              appliance.sourcegraph.com/managed: true

Data
====
spec:
----
spec:
  requestedVersion: "5.3.9104"

  blobstore: {}

  codeInsights:
    disabled: true

  codeIntel:
    disabled: true

  frontend:
    disabled: true

  gitServer:
    disabled: true

  indexedSearch:
    disabled: true

  indexedSearchIndexer:
    disabled: true

  pgsql:
    disabled: true

  postgresExporter:
    disabled: true

  preciseCodeIntel:
    disabled: true

  redisCache:
    disabled: true

  redisStore:
    disabled: true

  repoUpdater:
    disabled: true

  searcher:
    disabled: true

  symbols:
    disabled: true

  syntectServer:
    disabled: true

  worker:
    disabled: true

Events:
  Type    Reason             Age                From                   Message
  ----    ------             ----               ----                   -------
  Normal  ReconcileFinished  10s (x2 over 10s)  sourcegraph-appliance  Reconcile finished.