Design a data class as an interface for streaming as well as static (, historical) datasets

[BjoernLudwigPTB]

To make working with the agents easier, we should be able to clarify on how input data should be structured to best fit the agents design.

Regarding streaming data, we should as well keep in mind to adapt the generators of PyDynamic to this design and view this as very closely connected.

That instantly leads to an extended development of the signals themselves probably, which benefits PyDynamic as well, maybe in extending the previously started class Signal. Maybe we can even base the development on that class.

Originally posted by @BjoernLudwigPTB in #18 (comment)

[BjoernLudwigPTB]

With our use of time-series-metadata and our discussion in the Met4FoF project and the preparations for the SMSI web tutorials we have already taken some steps to resolving this. Basically what is needed now, is documentation.

[BjoernLudwigPTB]

Although we did not work on the interface between the agents and PyDynamic, the datastream topic is well handled in the agennts meanwhile. So I close this issue now, although there is still plenty of room for improvements. Just the original intend to think through the process of generating signals has actually been worked on sufficiently.