Markov kernel

I will be in need of something which captures linear and nonlinear Markov kernels.
Essentially parametrised measures `m(x)` which for fix `x` are just probability measures.
```
struct GaussKernel{T}
op::T
end
(k::GaussKernel)(x) = GaussMeasure(k.op(x))
```
with
```
k = GaussKernel(x -> (mu = 2*x + 1, Sigma = 1.0))
k(1.0) # GaussMeasure((mu = 3.0, Sigma = 1.0))
```
What do you think?