fix numkit float/int indexing

[orbeckst]

The following numkit tests in test_integration.py fail

• Test_simps_error.test_odd_const_x
• Test_simps_error.test_odd_const_x_dx

The problem is in numkit.integration.simps_error():

            if len(x.shape) == 1:
                shapex = numpy.ones(nd)
                shapex[axis] = x.shape[0]
                saveshape = x.shape
                returnshape = True
               x=x.reshape(tuple(shapex))

E               TypeError: 'numpy.float64' object cannot be interpreted as an index

The argument for reshap (shapex) must be an integer or tuple of integers.

Fix

Probably just

                shapex = numpy.ones(nd, dtype=np.int_)

Background

• see https://github.com/Becksteinlab/GromacsWrapper/pull/110#issuecomment-288475330 for initial discussion and e.g. https://travis-ci.org/Becksteinlab/GromacsWrapper/jobs/187547479 for failure in numkit test_integration
• see rbgirshick/py-faster-rcnn#481 for description: deprecated indexing with floats has been removed in numpy 1.12.

[dotsdl]

See #114.