feat: instance record management improvements

Previously, the instance resource was responsible for creating, updating
and deleting the instance record in Etcd. This led to a corner case in
our disaster recovery process when we need to remove an instance
resource without executing its lifecyce methods.

This commit shifts the responsibility for creating and deleting the
instance records to the workflows. Now, we will create or update the
instance record to indicate the operation we're about to perform.
We only completely delete the instance records if the database operation
was successful.

The instance resource still updates the instance record to indicate an
available or failed status with an error.

PLAT-398

Author: jason-lynch

api/apiv1/design/instance.go

@@ -162,6 +162,7 @@ var Instance = g.Type("Instance", func() {
 		g.Enum(
 			"creating",
 			"modifying",
+			"deleting",
 			"backing_up",
 			"available",
 			"degraded",