Extend functionality of@autoupdates

[bartvanerp]

The @autoupdates macro works great for streaming data, as long as it is used in very simple cases. I was trying to do a bit more complicated stuff and think I hit almost all possible cases where it fails:

• Functions in the macro cannot be nested, e.g. inv(cov(q(a)) does not work. So you will have to define foo(x) = inv(cov(x)), after which foo(q(a)) can be called.
• Conditional evaluations, including if-else statements are not possible. Again this can be circumvented by creating a separate function that includes this logic.
• Dealing with vectors of variables is ambiguous. As an example it is not possible to create for-loops over q(a) if a is a vector. Also indexing causes issues. Idem, broadcasting also does not work, e.g. mean.(q(a)) fails, whereas this actually seems necessary. For this last part it seems okay to create separate function again as broadcast_foo(...) = foo.(...). (Keep in mind here that some updates directly assign tuple, such as mean_var() and then this broadcasting needs to be more complicated)

[wouterwln]

I think this is related to #223 as a revamp of the rxinference functionality, so I'll schedule them for the same milestone