Simplify and speed-up math.hypot() and math.dist()

[rhettinger]

Apply the sumprod() DoubleLength functions to vector_norm() which is called by both hypot() and dist().

Baseline timing

% ./python.exe -m timeit -r11 -s 'from random import expovariate as r' -s 'from math import hypot' -s 'n=100' -s 'data = [r() for i in range(n)]' 'hypot(*data)' 500000 loops, best of 11: 579 nsec per loop 

Timing with PR

% ./python.exe -m timeit -r11 -s 'from random import expovariate as r' -s 'from math import hypot' -s 'n=100' -s 'data = [r() for i in range(n)]' 'hypot(*data)' 500000 loops, best of 11: 476 nsec per loop 

Timing with PR if fma() is not used

% ./python.exe -m timeit -r11 -s 'from random import expovariate as r' -s 'from math import hypot' -s 'n=100' -s 'data = [r() for i in range(n)]' 'hypot(*data)' 500000 loops, best of 11: 531 nsec per loop