Support kernels defined using keras functional API

[ojcobb]

Currently we call kernels as kernel(x, y) however models defined using keras's functional API require multiple inputs to be passed as a list. In these cases we instead need to call kernel([x, y]).

Ideally we would facilitate kernels that operate in either manner.

[ojcobb]

Particularly important given we require models defined using subclassing API to have the annoying get_config() and from_config() methods defined.