* allow the relationship's serializer key to be a Proc
* fixes
* specifically test the relationship name and the resource type
* support object blocks without a serializer defined
* stop validation gracefully if the relationship is polymorphic
* improve performance by using instance variables instead of accessor methods
* force initialization up front to avoid the iterative calls
* serializer procs should act like polymorphic when determining the id_method_name
* specs for serializer procs
* updated with more details and examples for relationship serializer options
* adjust specs to define the serializers
* avoid extra method calls for performance
* name change
* one less function call for better performance
* do not require lazy loaded relationships to resolve the serializer
* give polymorphic precedence for backwards compatibility
* move serializer inference into ObjectSerializer to allow for overriding
* move method for better diff
* fix race condition in multi-threaded environments
If you include a default empty `data` option in your JSON API response,
many frontend frameworks will ignore your `related` link that could be
used to load relationship records, and will instead treat the
relationship as empty.
This adds a `lazy_load_data` option which will:
* stop the serializer attempting to load the data and;
* exclude the `data` key from the final response
This allows you to lazy load a JSON API relationship.
This allows specifying a `:links` option to a has_many/has_one
relationship, which means you can specify `self` or `related` links as
per the JSON API spec (these are often useful for not loading all
associated objects in a single payload)
