[boost] MPL on IBM-vacpp

As you might have noticed we have succeeded in running regression.py on IBM. In the mean time Joaquin already noted that multi_index was broken. He also suggested that it was probably due to BOOST_MPL_CFG_USE_OPERATORS_OVERLOADING. And indeed, adding BOOST_WORKAROUND(__IBMCPP__, <= 600 ) to boost/mpl/aux_/config makes that multi_index compiles again (and probably other libs too). So is it OK to add the patch in attachment ?