NEW: MergeBeans function instead of updating existing bean from json

[nPraml]

When updating complex object graphs with the existing DB.toJson(existingBean, json) this might fail, if you have cyclic references.

Explanation
Ebean uses streaming when reading/writing jsons.
If you have a rather complex object graph with cycles, you may have problems when deserializing them.

As example take the class "User"

class User {
   @id int id;
   String name;
   @ManyToOne @WhoCreated User whoCreated;
   @ManyToOne @WhoCreated User whoModified;
} 

So your initial user could look like

User root = new User();
root.name = "Root"
root.whoCreated=root; // bootstrap ourself (makes cycle)
root.whoModified=root;
DB.save(root);

If you serialize this, your JSON will look like

{
  "id" : 1,
  "name" : "Root",
  "whoCreated": { "id" : 1 },
  "whoModified": { "id" : 1 },
}

Serializion works fine, it stores just references. When it comes to deserialization, the following happens:

bean1 = desc.createEntityBeanForJson();
bean1.id = 1;
bean1.name = "Root"
// first recursion
bean2 = desc.createEntityBeanForJson();
bean2.id = 1;
contextPutIfAbsent (first put)
bean1.whoCreated = bean2
// second recursion
bean3 = desc.createEntityBeanForJson();
bean3.id = 1;
contextPutIfAbsent (second put) - we will find "bean2" as contextBean
desc.merge(bean3, bean2) // copy all changed fields from bean3 to bean2 (there are none)
bean1.whoCreated = bean2

The more complex the json structure is, the more problems will appear.

This PR reverts the workarounds in ebean-orm#3000 (where we must buffer the json)
and mainly reverts ebean-orm#2617
(it will keep the API as deprecated)

It is recommeded now to do deserialize the JSON bean first and use Database.mergeBeans(jsonBean, targetBean)