optimize _calcSurplusAndCap() in LiquidationLibrary

[rayeaster]

address #683

[GalloDaSballo]

@dapp-whisperer Very delicate change

I added an assert so we can invariant test if that gets broken

We assume partial can never close a CDP, the assert ensures that
Due to the assert we can remove some of the unnecessary logic

Same idea for full liquidation, do the math only when needed

[GalloDaSballo]

On error: SL-05: The CDPs should be sorted in descending order of new ICR (accrued)

This is the issue we've already discussed plenty of times

[GalloDaSballo]

On broken Foundry Tests:
The new math formula causes overflow, I believe this happens due to the crazy values echidna uses

[GalloDaSballo]

Sent tweak with Assert so we can let echidna cook

[GalloDaSballo]

Waiting on @dapp-whisperer review before merging, we can differential fuzz if need be