Remove use of interaction_mask_indices on BG/Q
This field was degrading cache performance ~1% on x86. It probably
made little difference on BG/Q, because the extra integer operations
can use the second instruction-issue port, assuming the use of OpenMP
to use more than one hardware thread per core. Overall, this code is
about 1% faster on BG/Q.
Minor fix to the gmx_load_simd_4xn_interactions() function that looks
up the exclusion masks, so that new non-x86 platforms won't silently
fail for want of an implementation of this function.
Minor simplication to always pass simd_interaction_indices to
gmx_load_simd_4xn_interactions(), since it is only used on BG/Q and
then it is non-null.
Change-Id: I140a11607810e9cf08b702cae0b48426c3592fec