Add iterative refinement

[freemin7]

Iterative refinement is an algorithm for improving solutions.
The basic idea is that if you have an residual
r = b- Ax
You can calculate a correction c using the residual.
Ac = r
and x+c is an improved solution.
However this approach might run into floating point troubles, so there are version that are adapted for floating point arithmetic Lecture note 13 so this merits some investigation beyond the trivial implementation.