Multivar-horner

__________________

* added option `find_optimal` to find an optimal factorisation with A* search, explanation in readme
* optimized heuristic factorisation (more clean approach using just binary trees)
* dropped option `univariate_factors`
* added option `compute_representation` to compute the string representation of a factorisation only when required
* added option `keep_tree` to keep the factorisation tree when required
* clarification and expansion of readme and `example.py`
* explained usage of optional parameters `rectify_input=True` and `validate_input=True`
* explained usage of functions `get_gradient()` and `get_partial_derivative(i)`
* averaged runtime in speed tests

__________________

* introducing option to only factor out single variables with the highest usage with the optional parameter ``univariate_factors=True``
* compute the number of operations needed by the horner factorisation by the length of its recipe (instead of traversing the full tree)
* instead of computing the value of scalar factors with exponent 1, just copy the values from the given x vector ("copy recipe")
* compile the initial value array at construction time

__________________

* first stable release

__________________

* birth of this package