Add laws for MonadState and MonadReader

[Lysxia]

See #5. These two classes are the easiest to think about.

[Lysxia]

Ping @RyanGlScott

[RyanGlScott]

I really don't feel qualified to make a judgment about whether these should be the laws for these classes, which is why I haven't taken action. It might be worthwhile to bring up this topic on the Haskell libraries mailing list in order to garner a consensus (which I'd find convincing).

[Lysxia]

I see, thanks for your input! I will start a discussion on the mailing list.

[hanshoglund]

@Lysxia @RyanGlScott Any updates?

[Lysxia]

This is still WIP. On the mailing list it was suggested to investigate the matter more formally. So I started studying those laws in Coq: https://github.com/Lysxia/coq-mtl but I've been distracted by other things lately.

The situation, as I understand it, is that there is no formal definition of the mtl classes, and there is no objective criterion for what makes a law "good" or "bad". Through the formalization, the hope is to get a better understanding of how the laws relate to each other, and ideally to tease out some nice structure that could convincingly be considered the "essence" of mtl.

I'd be interested to hear about other people's motivations to have these laws documented.

[hanshoglund]

How about adopting the uncontroversial idea that reader, writer and state are monad homomorphisms as a starting point?

[Lysxia]

It's been a week since I reiterated my proposal: https://mail.haskell.org/pipermail/libraries/2019-October/030038.html

Is there anything more I can do to make progress to get this merged?