Out Of Bound optimising functionality about kerndisc

[BracketJohn]

oob functionaly in AS is implemented for each kernel and basically checks whether parameters of a kernel were optimized above or below a certain constraint for said parameter.

I should be able to implement this by either:

• doing manual checks on kernel params
• using some gpflow magic, maybe they have a way of not only assigning values, but also adding constraints

[BracketJohn]

After more research it seems like this is superfluous for our current goals. The checks can be (partly) summarized as:

        return any([self.mean.out_of_bounds(constraints), \
                    self.kernel.out_of_bounds(constraints), \
                    self.likelihood.out_of_bounds(constraints)])

(for mean, kernel, ll)

It seems like lloys/duvenauds optimizer was able to go out of bounds for model parameters. This alos applies to kernel params, e.g., they were able to move a CP out of X range.

This does not happen here, thus closing for now.