fixing the gradient optimization of IRR

SLSQP doesn't work with IRR. That's because in the cashflow method, the AEP is added as an int. This made the finite difference not work properly. Another fix of this branch is adding a dependency to numpy_financial, as the IRR function has moved from numpy to numpy_financial in the most recent versions of numpy.

Note that the IRR function was moved from numpy to numpy_financial in the most recent versions of numpy

@mmpe: I'm having trouble passing the tests, could you please help me out figuring what's going on? Thanks!

changed title from fixing the gradient optimization with respect to IRR to fixing the gradient optimization of IRR

It is three different things:

• The linux test fails because it uses the newest version PyWake which do not have the CubePower function in py_wake.windturbines
• the pycodestyle test fails because of some indent inconsistencies
• the windows test fails because a result value has changed. It will probably fail too, when PyWake is updated on the test machine or an update of pywake is added to the ci

closed