Jitcdde

* Added examples for networks and time-dependent delays to documentation.
* Improved error messages.
* Improved addressing of initial discontinuities in fringe cases such as ODEs.

* Added utility function `quadrature` for continuous delays.
* Some improvements of documentation.

* Fixing some under-the-hood inefficiencies.
* Hopefully fixed [this issue](https://github.com/neurophysik/jitcdde/issues/10) with Unicode and Jinja2 on MacOS.
* Some improvements to documentation.

There now is a SymEngine version number to refer to as a dependency – which means that automatically installing this dependency should work and you do not have to build SymEngine on your own anymore.

* Switched from SymPy to SymEngine as a backend, which may speed up code generation by a factor of up to nine hundred. Almost everything that worked before should still work. See [SymPy vs. SymEngine](http://jitcde-common.readthedocs.io/en/latest/#sympy-vs-symengine) for details. As a required [Symengine release is pending](https://github.com/symengine/symengine.py/issues/204), you have to [build it from source](http://jitcde-common.readthedocs.io/en/latest/#building-from-source) for now.
* Added OpenMP support.
* Added variant for transversal Lyapunov exponents.
* Several improvements to documentation.
* Changed some default parameters and behaviours.
* Dropped support of some old, rarely used syntax.
* Dropped Python 2 support.

* Utility functions for intialising the past from a function or as a constant.
* Some documentation improvements.
* Fixing some bugs with defining the past for `jitcdde_lyap`.

This will be probably the last release using SymPy for symbolics. The next one will use SymEngine.

This will also probably be the last release supporting Python 2.