feat: call webhooks at least once (#5454)

* feat: call webhooks at least once

* self review

* feat: improve notification observability

* feat: add notification tracing

* test(e2e): test at-least-once webhook delivery

* fix webhook notifications

* dedicated quota notifications handler

* fix linting

* fix e2e test

* wait less in e2e test

* fix: don't ignore failed events in handlers

* fix: don't ignore failed events in handlers

* faster requeues

* question

* fix retries

* fix retries

* retry

* don't instance ids query

* revert handler_projection

* statements can be nil

* cleanup

* make unit tests pass

* add comments

* add comments

* lint

* spool only active instances

* feat(config): handle inactive instances

* customizable HandleInactiveInstances

* call inactive instances quota webhooks

* test: handling with and w/o inactive instances

* omit retrying noop statements

* docs: describe projection options

* enable global handling of inactive instances

* self review

* requeue quota notifications every 5m

* remove caos_errors reference

* fix comment styles

* make handlers package flat

* fix linting

* fix repeating quota notifications

* test with more usage

* debug log channel init failures

cmd/defaults.yaml

@@ -171,6 +171,14 @@ Projections:
     Notifications:
       # As notification projections don't result in database statements, retries don't have an effect
       MaxFailureCount: 0
+    # The NotificationsQuotas projection is used for calling quota webhooks
+    NotificationsQuotas:
+      # Delivery guarantee requirements are probably higher for quota webhooks
+      HandleInactiveInstances: true
+      # As quota notification projections don't result in database statements, retries don't have an effect
+      MaxFailureCount: 0
+      # Quota notifications are not so time critical. Setting RequeueEvery every five minutes doesn't annoy the db too much.
+      RequeueEvery: 300s
 
 Auth:
   SearchLimit: 1000