[igor-sachok]

Implement widening for FPalt formats. For that:

1. add a typedef for alternative formats in spatz_pkg
2. add widening functions to spatz_pkg
3. change vfu: use the added functions based on source and destination formats to calculate the widened operands.
4. operands are widened to the destination format before reaching FPnew, so fpu_src_fmt must follow fm.dst, not fm.src, in the widening case.
   Note: Special operands (zero, subnormals, inf, NaN) are still not converted correctly by RTL. Alternative formats are implemented the same way as others.
   Note 2: In the current 32b configuration of the FPU, FP16alt and FP8alt are not enabled. That's why tests for them are under #if ELEN == 64.

The corresponding spatz_vpu branch can be found under the following link:
https://github.com/pulp-platform/spatz_vpu/tree/ihor/fix-vfu-fp-alt