support negative base in pow for integer exponent #88
Description
technically negative bases can be supported for integer exponents
currently negative bases simply error due to the log10 step
for example (-2)^2 = 4 and (-2)^3 = -8
for all even integer exponents (-a)^b = a^b
for all odd integer exponents (-a)^b = -(a^b)
the first issue is that non-integer exponents on a negative base can result in complex numbers, and so is categorically NaN in the float spec, even if there exists some answer
for example cube root of -8 is -2 but in javascript Math.pow(-8, 1/3) is NaN
so what we would want is to first check that the exponent is an integer and then check if it is odd/even to preserve or drop the - respectively
we don't currently have an isInteger but a.frac().isZero() should be a safe and relatively efficient check
determining odd/even is potentially more of an issue, the intuitive thing would be to do a.mod(2).isZero() but float modulo is an operation that has precision issues due to float representation limitations
for example 9.6%3.2 is 3.1999999999999993 rather than 0 in float math
the basic logic of % is x - floor(x / y) * y which produces the above precision issues
the question, that i don't currently know the answer to, is whether precision issues in float modulo ever actually manifest in decimal floats (vs. binary floats) where both x and y are integers (x is known to be an integer because that's the only exponents we support, and y is always 2)