ChebyCoeff::slope_a() type error leads to incorrect result

[jakelangham]

ChebyCoeff::slope_a() for nontrivial profiles typically returns extremely high values - at least for my build environment (gcc 6.4.0).

This is because the loop index n is given as auto, becoming an unsigned int, but its negation is required in the computation leading to an overflow.

I wonder if some compilers behave differently leading to this not being picked up before.

Anyway, this can be fixed quickly by using an int instead, as in ChebyCoeff::slope_b().