Small time steps when computating flow bounds

[pietercollins]

We currently often use very small steps when solving a differential equation. This is probably due to a poor implementation of the Bounder methods. We should fix this to allow reasonable steps.

[lgeretti]

As an alternative to the current integration procedure, we may use the approach in Flow*: compute the Taylor expansion of the flow up to a given order, then find a contraction from an initial guess on the remainder. It may fail, but then again from time to time we also fail finding a step size greater than the minimum allowed (or equivalently, our step size is so small that no significant progress is made).

In this case it depends on #325

[lgeretti]

I am going to add my implementation for bisection search of the maximum allowed step size to the EulerBounder, if it's ok with you @pietercollins . I will also use this occasion to introduce a bisection search class in the solvers to generalise the approach and improve readability.

[pietercollins]

We still are having issues with small time steps, and seriously need to address this better.

[lgeretti]

I am all for it, but do we have a proposal on how?