Adding Covariance Matrices

[BenSmithers]

Basic functionality is here - hoping to continue work on this on a branch in the IceCube PISA repo rather than my personal fork.

What's here, now

• Adding covariance matrices to do fits with correlated nuisance parameters
• Code to write and read functions to and from json files

What's not yet fully there

• Creating DerivedParams and defining functions in config files
• Serialization of the full state, including DerivedParams

[BenSmithers]

The last two commits addressed a few bugs that I came across while using these in some fits. DerivedParams are a bit closer to serializability now

[philippeller]

This looks great! So the old parameters still are usable from within each stage as if nothing happened, but internally they are represented in the rotated form, if I understood correctly?

[BenSmithers]

This looks great! So the old parameters still are usable from within each stage as if nothing happened, but internally they are represented in the rotated form, if I understood correctly?

That's right! It essentially turns the two Params into four. The original ones are replaced with DerivedParams whose values are functions of the new, rotated, ones.

[philippeller]

Cool! Is this PR thought as WIP, or do you want to get what you have merged?

[BenSmithers]

I'd like what I have here merged!

[philippeller]

cool! I then only want to run a few checks and if they pass merge it

[philippeller]

Run come additional checks, it all looks good and will merge now