Support for SIMD on X86 with VCL

[Idclip]

New Features:
    - OpenVDB now includes a copy of Agner Fog's VectorClass Library (VCL)
    for the use of generically wrapped SIMD intrinsics throughout the core
    library. VCL supports intel ISA up to AVX512. This support is OFF by
    default but can be enabled with USE_VCL=ON.

    - PointTransfer algorithms have been instrumented with SIMD instructions
    which result in significantly improved performance of anywhere up from
    2x to 4x. This includes all PCA and Point Rasterize methods (except for
    Ellipsoids).

Improvements:
    - The PointTransfer framework now allows transfer schemes to set a power
    of two Batch Size for bucket rasterizing. This results in improved
    performance for PCA and all PointRasterize methods, even without the use
    of VCL/SIMD intrinsics.