- Make existing `fast_interpolate` slightly faster with batch-inversion
  and Lagrange interpolation cutoff value
- Add new batch-version of fast_interpolate `batch_fast_interpolate` which
  also uses memoization. Up to 70x times faster than (the optimized)
  `fast_interpolate` in large batches.

Co-authored-by: Alan Szepieniec <alan@neptune.cash>