Remove unnecessary ICC flags affecting performance
Since we add -msse/../-mavx based on the acceleration we shouldn't
add -mtune=core2 anymore. Especially because it is added later
and takes precedence over the (higher) acceleration flag.
-ip and -funroll-all-loops could also be deleted because they don't
seem to give any significant performance improvement, and might
increase compilation time, but they don't hurt gromacs performance.
In theory it could help to use -xavx instead of -mavx but I can't
measure a difference.
Change-Id: Icd11c40c3cd3ef2ae6ef42f07d5d75c228593f51