Optimize all functions for 2D arrays

[tilleyd]

This change allows functions to be evaluated with different parameters simultaneously. As an example, if you have an array of shape (10, 50) containing 10 samples, previously you would have to evaluate them as

for i in range(10):
    y[i] = f1(x[i])

but can now be done simply with y = f1(x).

The functions have also been reimplemented using numpy array operations to allow a significant speedup for 2D inputs. For interest's sake, benchmark results measured with arrays of size (30, 100) (i.e. 30 samples of 100 dimensions) are listed below.

Function	Old Time (ms)	New Time (ms)
f1	1.4786	0.1097
f2	1.5524	0.1416
f3	1.8885	0.1353
f4	2.3877	0.1332
f5	2.3929	0.1261
f6	3.4209	0.1387
f7	3.7662	0.1435
f8	5.1008	0.1446
f9	5.4039	0.1465
f10	5.9024	0.1706
f11	5.7707	0.1838
f12	5.8063	0.2113
f13	5.7787	0.2223
f14	5.7659	0.2334
f15	5.6938	0.2403
f16	5.9140	0.2566
f17	6.4317	0.2809
f18	6.3171	0.2862
f19	6.7248	0.3025
f20	7.0393	0.3232
f21	7.3497	0.3329
f22	7.9632	0.3555
f23	8.7486	0.3813
f24	9.0502	0.3957
f25	9.5073	0.4114
f26	10.6782	0.4404
f27	11.7890	0.4701
f28	12.5931	0.4901
f29	13.3116	0.5293
f30	13.8205	0.5616