Allow assertions on mutable object properties.

[AndrolGenhald]

Follow up to #4619 to allow assertions on mutable properties as well.
All of the functionality is already there, we just have to not prevent it.

[AndrolGenhald]

Also, thanks @SCIF for the property assertion support, I hadn't noticed that until today!
I only just noticed that you excluded this due to references, I'll add some tests for that and see what happens.

[SCIF]

Hi @AndrolGenhald , thanks for this PR.

I have found another bug #6770 which, probably, is fixed by this PR, so could you please add tests of magic properties?

TBH, I didn't notice rememberPropertyAssignmentsAfterCall config directive at the time of initial implementation. But I was quite confused whether I should restrict the assertion scope with read-only properties and mentioned this in my initial PR. There was no a discussion about it so I designed a check this (pretty blunt) way. Honouring of rememberPropertyAssignmentsAfterCall would be much better way.

[AndrolGenhald]

@SCIF I'll take a look at that one too, currently digging through references and trying to add support for that. It doesn't look like Psalm really tracks references themselves yet, just whether or not they exist.

I'm not sure if rememberPropertyAssignmentsAfterCall should affect whether or not the assertions work. If the annotation is there I feel like it should work regardless, but I could see where it might be confusing. If rememberPropertyAssignmentsAfterCall === false does prevent the assertions from doing anything, the documentation should probably make that clear, or maybe it should just fail with the previous error you had, and change it to mention that config setting?

[AndrolGenhald]

After some thought, I think the solution I was working towards (adding a reference list to Union) isn't a good idea. I also thought about adding something to Context, but that's still going to require a lot of work to make sure the types stay in sync. I think the best solution is to refactor it so that references are stored differently, and actually point to the variable they reference somehow, but that's going to take a lot of work.

In the meantime, I'd prefer to just mark the failing tests as skipped and go ahead and get this merged. With how little most people use references I don't really expect it to be an issue in practice, even though it'd be nice if it worked correctly.

[weirdan]

Thanks!