Update to the new obake API and tbb API

This minor release updates and refreshes the code after one year of stable activity.

What's Changed

• update for tbb and obake. CI update by @darioizzo in #61

Full Changelog: v1.9.1...v1.9.2

Important BugFix release

As of v.1.7 the vectorized algebra contained a bug causing sometimes wrong results for vectorized gdual artihmetics at order >=2.
This release addresses that important bug and starts the 1.9.x numbering.

YACMA Update

This minor release only contains an update to the build system .

Pybind11

Moving to pybind11

• Following the recent trend of our toolchain, this release drops boost python in favour
  of pybind11.
• Some improvements are also made with the numpy interface adding increased interoperability

Obake

This is a new major release in which the computer algebra backend of Audi has been switched from piranha to obake (obake is piranha's successor project):

https://github.com/bluescarni/obake

As a result, Audi's performance has increased (up to 10x in some specific tests), and Windows is now officially supported. Audi now also requires a C++17 capable compiler.

One regression is that the pyaudi pip packages for Windows have been temporarily disabled due to the lack of support for MinGW by one of obake's dependencies. We are looking into re-enabling the pyaudi pip packages for Windows soon.

Python 37 support added

This release adds the python 37 support on the PyPi package manager.

Update third party dependencies

• This is a minor release including updates to soem of the third party dependencies we typically build our binaries with. In particular, linux binaries distributed via pip will now use boost 1.70 and Cmake 3.13.4

• Also the python test suite has been improved

Support for third part inclusion.

This release contains some small bug fixes allowing the headers to be included from third party projects without creating issues of multiple symbol definitions.

Updated build system

This release touches the build system only slightly as to allow OSX to link correctly to the python library.

Ordering gduals

The only change on this release, is the addition of an ordering for gduals, based on the implementation of the operators < and >. The ordering works comparing the constant coefficient of the gdual polynomial.