update ffp-model flag for new intel compiler

[acavelan]

The new intel compiler no longer supports "source" for the ffp-model. Using "strict" instead to enforce IEEE floating-point compliance.