Add worst-case option for BGV/BFV parameter generation

[yspolyakov]

For the expansion factor, we can use $N$ instead of $2 \sqrt{N}$ for the cases where worst-case estimates are desired.