CFW regression on Develop branch

[pedrocamargo]

Purpose of use
Development

Describe the bug
CFW is no longer working, and is causing tests to fail

To Reproduce
See #283

[janzill]

To summarise the discussion in #283, whitespace changes in bpr.pyx led to the bug. Changing it back slightly altered traffic assignment test results, see #293. I'm removing the labels here (bug and critical) because they do not apply anymore but I'm keeping this issue open for now because I want to understand what is happening. It could be that very small changes in results (numerical precision) lead to larger differences in the step finding routine.

[janzill]

Things to look at:

• making sure we use double precision in calculations everywhere
• compare generated c++ code before and after white-space changes in bpr.pyx
• investigate step finding optimisation parameters