plan_fft should be more generic

The planning functions (`plan_fft`, `plan_fft!`, `plan_ifft`, etc) are currently derived from FFTW. The docs and method signature refer to library-specific options: the planning modes (e.g., FFTW.MEASURE) and timelimit. As I understand it, these are not universal standards: MKL can preallocate output and has other options (DFTI_ORDERING) and other back-ends may do something different---or even nothing at all.

Consequently, it might be clearer to describe the planning functions here as `plan_fft(A, [dims]; kwargs...)` and have FFTW.jl provide `plan_fft(A, [dims]; flags=FFTW.ESTIMATE, timelimit=Inf)`.

This would not require any changes to the code but would require a concurrent change to the FFTW.jl docs. It already includes docs for some functions not represented in AbstractFFTs, so this seems feasible.