Speed up lagrange in pallas_based_plonk

[volhovm]

Right now lagrange funciton is doing

Array.init domain_size ~f:(fun i ->
    Kimchi_bindings.Protocol.SRS.Fq.lagrange_commitment srs domain_size i )

which is calling the bindings function as many times as the domain size. This is concretely expensive: 150ms for just querying all the lagrange bases.

Alternative suggestion:

1. Implement this function to query lagrange bases directly from rust side using 1 FFI call (this is already very fast)
2. Do not read it from the SRS every time, but instead read once and pass as arguments to necessary functions