Add arora2013 solver

[jorgepiloto]

Solves for #3 by adding arora2013 solver.

[jorgepiloto]

The original report contains a critical error in the equations defining the way final velocity vectors are built from f and g functions. In addition, some signs concerning the W variable and its derivatives seem to be wrong.

At the moment, only the zero-revolution is implemented with all the test cases passing. Previous singularities were all commented within the source-code.

[jorgepiloto]

The complexity of the algorithm for the multi-revolution case is high. Because the timeline for the first release approaches and the zero-revolution case is heavily tested, let us implement this partially, discarding for the moment the multi-revolution scenario.

[jorgepiloto]

Before merging this, I would like to point it out the necessity of increasing the coverage of this algorithm. Even if the test-suite included at the moment might be seen robust, the fact that Arora's routine implements so many conditionals to check for solution region decreases the overall one. Anyway, package coverage remains over 90% at the moment, so merging this.