[ENH] kernel mixture distribution

[fkiraly]

For distribution estimation, an important distribution type are kernel mixture distributions.

An open design question is how to represent these - Mixture allows mixture of arbitrary distribution objects, but:

• not all current kernels are distributions or can be currently represented as distirbutions
• there are no objects in skpro for kernels a the moment, so there is an open desigh question how to represent these
• parameterization is unclear - most common in literature is kernel choice, kernel parameters, and points of support

Compare sklearn KernelDensity - https://scikit-learn.org/stable/modules/generated/sklearn.neighbors.KernelDensity.html#sklearn.neighbors.KernelDensity