Explore alternative refinement strategies

[Luiskitsu]

In the current version, morphs in the chain are refined sequentially, assuming parameter independence. However, in practice, some morphs like squeeze and funcy may be correlated, which may cause instability depending on initial conditions.
Proposed ideas to test:

• Try alternative optimizers such as differential evolution or Basin-hopping
• Try simultaneous refinement of morphs rather than sequential refinement
• Change the order of morphs in the chain to test for independence
• Mimick a rietveld-style approach where a few parameters are refined first to get good starting values, additional parameters are gradually turned on and all the parameters are refined at the end.