Normalize YAML manifests before running diff

The current behavior of helm-diff is to show diffs even when the
YAML files are semantically identical but they contain style
differences, such as indentation or key ordering.

As an example, the following two YAMLs would be showing diffs:
```
apiVersion: v1
kind: Service
metadata:
  name: app-name
  labels:
    app.kubernetes.io/name: app-name
    app.kubernetes.io/managed-by: Helm
spec:
  ports:
    - name: http
      port: 80
      targetPort: http
  selector:
    app.kubernetes.io/name: app-name
  type: ClusterIP
```
```
apiVersion: v1
kind: Service
metadata:
  labels:
    app.kubernetes.io/managed-by: Helm
    app.kubernetes.io/name: app-name
  name: app-name
spec:
  ports:
  - name: http
    port: 80
    targetPort: http
  selector:
    app.kubernetes.io/name: app-name
  type: ClusterIP
```
even though they are semantically the same.

This commit exploits the reordering and normalization that is performed
by the yaml package when marshaling by unmarshaling and re-marshaling
all manifests before running the diff, thus eliminating the issue.

Author: tommasopozzetti

manifest/parse.go

@@ -135,7 +135,7 @@ func parseContent(content string, defaultNamespace string, excludedHooks ...stri
 
 	if parsedMetadata.Kind == "List" {
 		type ListV1 struct {
-			Items []yaml.MapSlice `yaml:"items"`
+			Items []map[interface{}]interface{} `yaml:"items"`
 		}
 
 		var list ListV1
@@ -163,6 +163,16 @@ func parseContent(content string, defaultNamespace string, excludedHooks ...stri
 		return result, nil
 	}
 
+	var object map[interface{}]interface{}
+	if err := yaml.Unmarshal([]byte(content), &object); err != nil {
+		log.Fatalf("YAML unmarshal error: %s\nCan't unmarshal %s", err, content)
+	}
+	normalizedContent, err := yaml.Marshal(object)
+	if err != nil {
+		log.Fatalf("YAML marshal error: %s\nCan't marshal %v", err, object)
+	}
+	content = string(normalizedContent)
+
 	if isHook(parsedMetadata, excludedHooks...) {
 		return nil, nil
 	}