Csaps

* Introduced optional `normalizedsmooth` argument to reduce dependence on xdata and weights [47](https://github.com/espdev/csaps/pull/47)
* Update numpy and scipy dependency ranges

* Bump numpy dependency version

* Bump scipy dependency version
* Bump sphinx dependency version and use m2r2 sphinx extension instead of m2r
* Add Python 3.9 to classifiers list and to Travis CI
* Set development status classifier to "5 - Production/Stable"
* Happy New Year!

* Fix using 'nu' argument when n-d grid spline evaluating [32](https://github.com/espdev/csaps/pull/32)

* Fix n-d grid spline evaluating performance regression [31](https://github.com/espdev/csaps/pull/31)

* Use `PPoly` and `NdPPoly` base classes from SciPy interpolate module for `SplinePPForm` and `NdGridSplinePPForm` respectively.
* Remove deprecated classes `UnivariateCubicSmoothingSpline` and `MultivariateCubicSmoothingSpline`
* Update the documentation

**Notes**

In this release the spline representation (the array of spline coefficients) has been changed
according to `PPoly`/`NdPPoly`.
See SciPy [PPoly](https://docs.scipy.org/doc/scipy/reference/generated/scipy.interpolate.PPoly.html)
and [NdPPoly](https://docs.scipy.org/doc/scipy/reference/generated/scipy.interpolate.NdPPoly.html) documentation for details.