Method push_forward for random variables

[JonathanWenger]

Is your feature request related to a problem? Please describe.
When sampling from random variables it is sometimes convenient to provide samples from a base measure and then transform them.

Give an example use case.
Creating a visualization of Gaussian process samples which move continuously.

Describe the solution you'd like.
A method of the form below or similar.

def push_forward(sample : np.ndarray, measure : str)
  return self.mean + self.cov.cholesky() @ sample

Note:
Consider different naming in #344, where @pnkraemer uses transform_base_measure_realizations(base_measure_realizations, locations, size) to implement this pattern for RandomProcess Posteriors.

[JonathanWenger]

An interface for this function may be added in #387.