Updates #30

Fixing typo

The arbitrary precision Sqrt implementation here was committed into
the standard library as big.Float.Sqrt.

This change deletes the old implementation and switches to using the
stdlib one in the other methods.

name          old time/op    new time/op    delta
Log/100-4        141µs ±29%      82µs ±24%  -41.89%  (p=0.008 n=5+5)
Log/1000-4       488µs ±16%     306µs ±24%  -37.25%  (p=0.008 n=5+5)
Log/10000-4     6.45ms ±26%    5.73ms ±30%     ~     (p=0.548 n=5+5)
Log/100000-4     142ms ±12%     132ms ± 4%     ~     (p=0.056 n=5+5)

name          old alloc/op   new alloc/op   delta
Log/100-4       41.6kB ± 0%    24.9kB ± 0%     ~     (p=0.079 n=4+5)
Log/1000-4       128kB ± 0%      75kB ± 0%  -41.34%  (p=0.016 n=4+5)
Log/10000-4     1.60MB ± 0%    1.22MB ± 0%  -23.41%  (p=0.008 n=5+5)
Log/100000-4    21.5MB ± 0%    18.7MB ± 0%  -12.90%  (p=0.008 n=5+5)

name          old allocs/op  new allocs/op  delta
Log/100-4          398 ± 0%       143 ± 0%  -64.07%  (p=0.008 n=5+5)
Log/1000-4         804 ± 0%       243 ± 0%  -69.78%  (p=0.008 n=5+5)
Log/10000-4      1.50k ± 0%     0.40k ± 0%  -73.38%  (p=0.008 n=5+5)
Log/100000-4     2.53k ± 0%     0.72k ± 1%  -71.75%  (p=0.008 n=5+5)

Pow(-2, 3) previously panicked with "Pow: negative base", even
though (-2)^3 = -8 is well-defined. This matches the behavior of
math.Pow from the standard library.

Now handles negative base when the exponent is an integer by
computing |z|^w and negating the result for odd exponents.
Still panics for negative base with non-integer exponent (which
would require complex numbers).

Fixes #35