Migrate use of Expect test patdiff to Myers

[mbarbin]

No description provided.

[mbarbin]

cc @tmattio - This PR experiments with the new windtrap.myers package.

Diff looks good! There are a few cases of an empty diff whose rendering has become a bit more verbose now. See in the PR review files:

Empty diffs used to be rendered like this (empty):

  [%expect {||}];

They are now rendered like this:

  [%expect
    {|
    --- expected
    +++ actual
    |}];

I think the empty output is probably a more ergonomic default? Probably also what diff does when run from the terminal. Could that be a good request to submit to the myers packages?

[tmattio]

Nice!!
Yes I agree, empty default is much better, generally, I'm keen to get closer to ppx_expect, although I found that LLM sometime get confused with what is expected vs actual, so the hint might be an acceptable divergence on non-empty diff

[mbarbin]

Hi! I didn't think much about the cases where the diff is non-empty, I don't feel strongly. I don't think I would mind that divergence indeed. Thanks for looking! I have open a few PRs like this one in a few projects, I have to go offline for a bit, but I'll send you soon a list if you'd like to look just for reference in case there's something interesting to note, but from what I can tell, all is working well. Thanks!

[mbarbin]

Hi @tmattio so I ended up going the vendoring route, and carry some minimal changes to make the integration easier in the tests. So far I have 4 projects where I'd end up doing this. Something doesn't really feel right to me, so maybe in a little while I'll try requesting again (as a gentle insistance) that you'd release myers as a standalone package! ( 😄 ). Team Vendor it is for now! Thanks again for windtrap!