Speed up correlation computation using SIMD instructions

Speeds up the computation of correlation values in
  wb_command -cifti-correlation,
  wb_command -cifti-correlation-gradient, and
  wb_view's dynamic connectivity feature
using SIMD-based dot product implementations.

To achieve this, the two modules cpuinfo and dot, from
  https://github.com/kloewe/mod-cpuinfo/commit/3799c70 and
  https://github.com/kloewe/mod-dot/commit/3b83d6a, respectively,
are used.

Depending on the definition of CARET_DOTFCN, src/Common/dot_wrapper.h
includes the functions from the dot module or an inline function that
implements the originally used dot product computation. If the dot module
is used, the function sddot() dispatches the call to sddot_avx(),
sddot_sse2(), or sddot_naive(), depending on the supported instruction set
extensions. These are determined at runtime using the functions from the
cpuinfo module.

The new functions are currently only used in the Linux version.