Poor numerical conditioning for large Lipschitz bounds

When using `LipschitzRENParams`, it stands to reason that the REN behaviour should converge to `ContractingRENParams` if the Lipschitz bound is very large. However, the parameterisation we use means that `D22` becomes extremely large too. 
- Can this be solved by some neat programming trick?
- Do we need to re-parameterise `D22` in another way?