🐛 Fix handling of new symmetric add operation

[michael-platzer]

I'd like to apologize, when submitting PR #114 adding the new symmetric add operation ADDS, I overlooked three things:

• the new operation of course needs to be added to the case statements that assign default values to the operands based on the operation in both fpnew_fma.sv and fpnew_fma_multi.sv
• the multi FMA's addend exponent needs to be rebiased for the dst format (so far this was not necessary because the addend was already in dst format - however, for the ADDS operation it is not)
• when extracting an FMA lane's local operands the operands_i[2] needs to be shifted using the correct format, which for the ADDS operation is the src format and not the dst format

This PR fixes these issues. Please excuse my sloppiness 🙈