Pink-noise-rl

New functionality:
- Vectorized sampling of more than one time step is now possible with `ColoredNoiseProcess.sample(T)`
- `ColoredNoiseProcess` now also has a dedicated `reset` function for refilling the buffer

This version introduces some breaking changes. It is no longer necessary to pass a list of beta values of the same dimensionality as the action space. Instead, a single beta can be passed, in addition to the action space dimensionality, and the noise will be sampled in a vectorized manner (in contrast to the iterative implementation in version 1).