Add get lde functions

[sai-deng]

I have to revert #91 to align the value domains.

[codeblooded1729]

I tried to work out why the mismatch was happening, and it seems that it was due to the following.
Basically the quotient polynomial constraint looks like
$$(x^n - 1) q(x) = VanishingPoly(x)$$
Due to the $x^n$ term, it becomes critical, over which coset of natural domain we would perfrom LDE (and commit) $q(x)$ on. That is we need to know the coset shift used (since $x^n$ evaluates to different values for different coset shift)
I see that current code performs LDE with constant coset shift given by F::coset_shift(), for all polynomials used. I am fine with this revert for now. Maybe in future we can try incorporating using custom coset shift into the logic used for quotient polynomials.