Support >= 3 dimensional generic ffts

[hsgg]

I recently had the need (or, rather, I thought I had the need) for arbitrary precision multi-dimensional FFTs. So I implemented them here. I also took the chance to optimize the power-of-two case a little.

It gets within a factor ~4 of FFTW without wisdom:

$ julia --project=GenericFFT -t 8

julia> A = randn(ComplexF64, 512, 512, 512);

julia> using GenericFFT, FFTW

julia> FFTW.set_num_threads(8)

julia> @time fft(A); @time fft(A);
  0.892791 seconds (2.31 k allocations: 2.000 GiB, 9.39% gc time, 2.88% compilation time)
  0.853025 seconds (128 allocations: 2.000 GiB, 4.21% gc time)

julia> @time GenericFFT.generic_fft(A); @time GenericFFT.generic_fft(A);
  3.884851 seconds (3.94 M allocations: 4.198 GiB, 0.92% gc time, 108.37% compilation time)
  3.253270 seconds (132 allocations: 4.000 GiB, 1.83% gc time)

which was on a MacOS M1:

julia> versioninfo()
Julia Version 1.11.5
Commit 760b2e5b739 (2025-04-14 06:53 UTC)
Build Info:
  Official https://julialang.org/ release
Platform Info:
  OS: macOS (arm64-apple-darwin24.0.0)
  CPU: 10 × Apple M1 Max
  WORD_SIZE: 64
  LLVM: libLLVM-16.0.6 (ORCJIT, apple-m1)
Threads: 8 default, 0 interactive, 4 GC (on 8 virtual cores)