Add FP8 and FP8ALT support to THMULTI DivSqrt
- This PR adds
FP8andFP8ALTsupport when selectingTHMULTIDivSqrt (building on top of #131). - Since the
THMULTIDivSqrt unit leverages a high-precision datapath, smaller LUT-based DivSqrt units supportingFP8andFP8ALTare instantiated for 8-bit-wide SIMD lanes.