lib: fix Windows parameter handling in LUT function

The 16x8-bit parallel lookup functions have two parameters:
a pointer to the LUT and __m128i containing the indices.
On Linux, this __m128i is passed through xmm0, but on Windows,
this is stored in the stack and a pointer to this address
is passed through a GP register.
Therefore, this memory must be read into xmm0, so it is compatible
with the Linux implementation.

This commit fixes issue #59.