fftw add sse sse2 avx avx2 options

[bvnp44]

Package and Environment Details (include every applicable attribute)

• Package Name/Version: fftw/3.3.8
• Operating System+version: Linux Ubuntu 16.04
• Compiler+version: gcc-5.4
• Docker image: conanio/gcc7
• Conan version: Conan 1.11.0
• Python version: Python 3.7.0

Description of your feature

Library support compilation with this options, you can see this in their CMakeLists.txt, but all of them are disabled by default, will be good to have them in conanfile.py

option (BUILD_SHARED_LIBS "Build shared libraries" ON)
option (BUILD_TESTS "Build tests" ON)

option (ENABLE_OPENMP "Use OpenMP for multithreading" OFF)
option (ENABLE_THREADS "Use pthread for multithreading" OFF)
option (WITH_COMBINED_THREADS "Merge thread library" OFF)

option (ENABLE_FLOAT "single-precision" OFF)
option (ENABLE_LONG_DOUBLE "long-double precision" OFF)
option (ENABLE_QUAD_PRECISION "quadruple-precision" OFF)

option (ENABLE_SSE "Compile with SSE instruction set support" OFF)
option (ENABLE_SSE2 "Compile with SSE2 instruction set support" OFF)
option (ENABLE_AVX "Compile with AVX instruction set support" OFF)
option (ENABLE_AVX2 "Compile with AVX2 instruction set support" OFF)

option (DISABLE_FORTRAN "Disable Fortran wrapper routines" OFF)

[thebrandre]

Closely related to that: it would be nice to support multiple precision versions as described here in the official documentation.

Fedora packages for fftw, for example, offer that kind of support for float, double, long double, and quad precision. This requires multiple runs of cmake, though.