Have a `logpdf` and `loglikelihood` for `NamedTuple` of f and y

[theogf]

Right now to compute logpdf or loglikelihood given a pair of GP samples f and observations y one needs to do:

logpdf(likelihood(f), y)

would that make sense to have a small wrapper

function logpdf(like::Likelihood, f_y::NamedTuple)
  logpdf(likelihood(f_y.f), f_y.y)
end

since this is what rand would return for a LatentGP :
https://github.com/JuliaGaussianProcesses/AbstractGPs.jl/blob/e7c67bd24c454f0b8179032ecbb82284887159c0/src/latent_gp.jl#L37

(and similarly for loglikelihood)

[st--]

I can see the appeal, but then I think we should have a proper type for this. Dispatching on generic NamedTuple without specifying what fields that tuple might contain seems like it could easily lead to bugs somewhere down the line.

[theogf]

We can dispatch on NamedTuple{(:f, :y)}