Instance Annotations in Atom should use metadata:Annotation element

OData supports applying annotations to data instances according to each payload.

In the atom format, we currently specify that an annotation is any element in a namespace other that the Atom, OData data:, or OData metadata: namespace with a "target" attribute. Properties of the annotation are defined in this same namespace.

This mechanism requires defining namespaces for each set of terms which include the values and the properties of the values. On the client it requires understanding parsing each element to determine whether or not it has the special "target" attribute indicating it should be interpreted as an annotation. When interpreting the value of a structured annotation, the parser has to understand properties in this special namespace, rather than the odata data: namespace used to parse non-annotation properties.

It would be all together easier if we defined an Annotation element in the OData metadata: namespace. This would make it easy for parsers that wanted to identify, or ignore, annotations to do so. Properties within structured annotations should be in the OData data: namespace, so that parsers could interpret the data within a structured annotation the same as any other structured value.