Fixed parameter overflow in divsqrt_multi

[sifferman]

Hello!

Issue

Commits 51e6ba8 and 1548dae from PR #102 have introduced a bug due to the use of NUM_INP_REGS-1 without checking NUM_INP_REGS > 0.

This causes the basic instantiation example to fail in DC due to Error: fpnew_divsqrt_multi.sv: integer overflow occurs. (ELAB-210).

Fix

I now split all 3 references to reg_ena_i[NUM_INP_REGS-1] into a new signal last_reg_ena that is forced to 1'b0 if NUM_INP_REGS==0.

Now the basic instantiation example once again is synthesizable.

Thanks!

[pascalgouedo]

PULP DIV/SQRT not used in CV32E40P.

Letting people using it approve it or not...