Compatibility with wolf::Scalar = float is broken

Created by: joansola

We designed wolf::Scalar so that, one day, we could just compile everything with a different scalar type, e.g., float. This could reveal interesting when trying to make WOLF portable to embedded devices that might be running smaller CPUs. I am not sure that this is really useful, but still we thought it was cool to have.

Now this compatibility is broken, because part of the API uses double, especially for issues that relate WOLF to outer libraries such as Ceres.

This is not a severe issue. If one day we are in the mood, we can fix it:

• the code that needs to be compatible with fixed types will have to use casts such as (Scalar)xxx , (Scalar*)xxx or whatsoever.
• This job is a pain...