Suppress nullable property/field initialization warning for required members

[333fred]

Test plan #57046

[333fred]

@RikkiGibson @jcouv for review.

[RikkiGibson]

Consider testing a constructor with [SetsRequiredMembers].

[333fred]

This isn't implemented yet. That's next PR. However, SetsRequiredMembers will not affect behavior here: LDM made the decision that SetsRequiredMembers does not ensure that required members are set in the constructor, so no nullable warnings should be given either.

[RikkiGibson]

That's fine. Let's just add the test at the appropriate time to ensure the expected behavior is recorded.

[RikkiGibson]

Was it considered, instead of skipping checking the flow state on exit, to behave as though the property is being assigned with something "valid" at the exit point? Curious if there's appetite to support scenarios like the following:

class C
{
    private string _field;
    public required string Field { get => _field; [MemberNotNull(nameof(_field))] set => _field = value; }

    public C() { } // no warnings
}

It's fine if this isn't of interest to the language team, or if it's a "stretch goal" so to speak that might get added later as time permits, but thought I would check.

[RikkiGibson]

Also curious about what should happen in this scenario:

public class C
{
    public required string Prop { get; set; }

    public C(bool unused) { }

    public C() : this(true)
    {
        Prop.ToString(); // warning?
    }
}

[jcouv]

Added Area-Compilers label (that's why I hadn't noticed this PR)

---

In reply to: 1077939493

[333fred]

@RikkiGibson please take another look.

[333fred]

You may want to review this section with whitespace turned off.

[333fred]

@RikkiGibson @jcouv for another look please.

[jcouv]

LGTM Thanks (iteration 3). Adding comment would be appreciated. Nits up to you.

[333fred]

I'll address the nits in the next PR