writeType and writeObject in SparkR should be handled by S3 methods

Corollary of https://issues.apache.org/jira/browse/SPARK-28040

The way writeType and writeObject are handled now feels a bit hack-ish, would be easier to manage with S3 or S4.

NB: S3 will require changing the order of arguments to call classes on object (current first argument is con)