Multivar-horner

__________________

ATTENTION: API changes:

* removed ``validate_input`` arguments. input will now always be validated (otherwise the numba jit compiled functions will fail with cryptic error messages)
* black code style
* pre-commit checks

__________________

Post-JOSS paper review release:

* Changed the method of counting the amount of operations of the polynomial representations. Only the multiplications are being counted. Exponentiations count as (exponent-1) operations.
* the numerical tests compute the relative average error with an increased precision now

__________________


ATTENTION: API changes:

* ``TypeError`` and ``ValueError`` are being raised instead of ``AssertionError`` in case of invalid input parameters with ``validate_input=True``
* added same parameters and behavior of ``rectify_input`` and ``validate_input`` in the ``.eval()`` function of polynomials


internal:

* Use ``np.asarray()`` instead of ``np.array()`` to avoid unnecessary copies
* more test cases for invalid input parameters

__________________

* BUGFIX: factor evaluation optimisation caused errors in rare cases. this optimisation has been removed completely. every factor occurring in a factorisation is being evaluated independently now. this simplifies the factorisation process. the concept of "Goedel ID" (=unique encoding using prime numbers) is not required any more
* ATTENTION: changed polynomial degree class attribute names to comply with naming conventions of the scientific literature
* added __call__ method for evaluating a polynomial in a simplified notation ``v=p(x)``
* fixed dependencies to: ``numpy>=1.16``, ``numba>=0.48``
* clarified docstrings (using Google style)
* more verbose error messages during input verification
* split up ``requirements.txt`` (into basic dependencies and test dependencies)
* added sphinx documentation
* updated benchmark results

tests:

* added test for numerical stability
* added plotting features for evaluating the numerical stability
* added tests comparing functionality to 1D ``numpy`` polynomials
* added tests comparing functionality to naive polynomial evaluation
* added basic API functionality test

internal:

* added class ``AbstractPolynomial``
* added typing
* adjusted publishing routine
* testing multiple python versions
* using the specific tags of the supported python version for the build wheels
* removed ``example.py``

__________________


* NEW FEATURE: changing coefficients on the fly with ``poly.change_coefficients(coeffs)``
* NEW DEPENDENCY: ``python3.6+`` (for using f'' format strings)
* the real valued coefficients are now included in the string representation of a factorised polynomial
* add contribution guidelines
* added instructions in readme, ``example.py``
* restructured the factorisation routine (simplified, clean up)
* extended tests

__________________

* support of newer numpy versions (ndarray.max() not supported)
* added plotting routine (partly taken from tests)
* added plots in readme
* included latest insights into readme