Discrete transforms need their x-value shifted

[JeffreyEarly]

The discrete cosine/sine/fourier transformations need to use (xVal - x.domainMin) for the xVal. This bug unfortunately passed my unit tests because I was testing on an unshifted domain.

[JeffreyEarly]

Actually, for best precision, we need to just implement the discrete form that doesn't depend on the dimension values (too many floating point multiplications). I fixed the cosine already.