Implement missing ARM NEON vector types

[gnzlbg]

Some of the ARM NEON intrinsics require the following types that are not currently available:

• f16, f16x4, f16x8: half-floats
• p8x8, p8x16, p16x4, p16x8, p64, p64x1,p64x2, p128: polynomial coefficient types (should maybe either be aliases to u... or newtypes of u...).

Given that the F16C instruction (_mm_cvtps_ph, _mm_cvtph_ps, ...) set also supports half-floats on x86 we probably want to provide some of the half-float types for x86 as well.

[gnzlbg]

Closed in #384 . The only types missing are the f16 vector types covered in #344 and the more convoluted float32x4x2_t structs covered in #316 .