x86: templatize SIMD FP arithmetic templates

Reduce redundancy, in preparation of the addition of further counterparts
for AVX10.2. Provide the "ne" parameter needed there right away, even if
unused for now.