Add support for Beta distribution in Generalised Bayesian Filters nodes

[LegrandNico]

Generalised Bayesian filtering is described in this paper.

We need to implement a node that uses:

• A fixed learning rate
• A dynamic learning rate (i.e. using continuous nodes to filter sufficient statistics)

The first steps will be to:

• Add a Beta class in the math.py file to convert between the different components of a Beta distribution.

See this page for the details on the parameterisation of a beta distribution, and this page can also be useful.