Fix CI, require AbstractMCMC 5.10, and allow MCMCChains 7

[astrozot]

MCMCChains version 7 introduces very minor breaking changes that do not affect the creation of chains, but just their display. Therefore it should be included in the compat entry. CI was red before this bump too, so the branch also unbreaks it: the dead actions/cache@v1 that auto-failed every test job, the missing stdlib compat entries and the step piracy that Aqua flagged, and the docs build, which needed nested_isdone to stop treating AbstractMCMC 5.10's AbstractProgressKwarg as a boolean. That last one is breaking, so this is released as 0.10.0 with a new HISTORY.md, dropping AbstractMCMC 3 and 4 and raising the Julia floor to 1.10. Verified locally against AbstractMCMC 5.16 and MCMCChains 7.7.

[shravanngoswamii]

MCMCChains is archived now, so better use FlexiChains, I will update this package and fix the CI, happy to merge this PR than. Thanks!

[astrozot]

@shravanngoswamii thank you for your kind reply and quick reaction! I saw the archived message on top of MCMCChains GitHub page, but did not know about the new FlexiChain. Is that the official replacement for MCMCChains? I do not see any explanation in the Readme of MCMCChain...

[shravanngoswamii]

Is that the official replacement for MCMCChains?

Yes, for now it's official but it's not maintained by TuringLang or chalk-lab maintainers. It's maintained by one of our past maintainers on her personal account.

I do not see any explanation in the Readme of MCMCChain

Ohh, yeah it was only updated in Turing docs. MCMCChains is not being developed further now, but it's still usable. FlexiChains is better alternative.

[shravanngoswamii]

Thank you @astrozot!