v3.0: bfloat16 mutiplier + adder

Releasing the fully tested, performance optimized bfloat16 multiplier using round toward zero rounding.
Once more all $2^{32}$ input combination have been tested, with the results validated against a golden model.

Multiplier features (new):

• classic high performance multiplier design
• assumes synthesized will be expand booth multiplier placeholder module to a 8b input radix 4 unsigned booth multiplier (implementation defined)
• round towards zero rounding mode
• no denormal support: rounded to $\pm 0.0$ in testbench, conserving the infinite precision's value's sign
• no support NaN and inf values as input: rounded to $+0.0$ in testbench

Adder features:

• dual path architecture
• round towards zero rounding mode
• no denormal support: rounded to $\pm 0.0$ in testbench, conserving the infinite precision's value's sign
• no support NaN and inf values as input: rounded to $+0.0$ in testbench

Misc:

• reworked C++ dpi and SV testbench code to share common code between adder and multiplier testbenches

For additional information, check the README.md at tag v3.0.