PERF: Use numpy arrays in andrews_curves plots

[khs26]

Following on from discussion in #11534, this uses a list comprehension in the following function:

def function(amplitudes):
    def f(t):
        x1 = amplitudes[0]
        result = x1 / sqrt(2.0)
        harmonic = 1.0
        for x_even, x_odd in zip(amplitudes[1::2], amplitudes[2::2]):
            result += (x_even * np.sin(harmonic * t) +
                       x_odd * np.cos(harmonic * t))
            harmonic += 1.0
        if len(amplitudes) % 2 != 0:
            result += amplitudes[-1] * np.sin(harmonic * t)
        return result
    return f

As both amplitudes and t are numpy ndarrays, there is probably some performance gain to be had from vectorising the loop.

My numpy knowledge and available time aren't sufficient to make this entirely trivial. I came across the following issues in particular:

• Basic slices of numpy ndarrays return views rather than copies, which then can't be subject to many numpy operations (i.e. np.resize).
• Addressing amplitudes in the inner scope assigns the name in the inner scope (see here for details). This causes an amplitudes is not defined exception.

[jreback]

see if you can provide the proper input above, so one can simply copy-paste to test this.

[khs26]

As per further comments in #11534, I've had a go and posted timings.

[sinhrks]

Closed by #11534. Pls reopen otherwise.