Partial properties: diagnostics for mismatch between parts

[RikkiGibson]

Test plan: #73090

[RikkiGibson]

The goal of this change is to make it so the getter symbol in public virtual T Prop => ... is not treated as introducing any of its own modifiers.

This makes the behavior of APIs like LocalAccessibility more consistent with properties declared using the accessor list syntax public virtual T Prop { get => ... }.

[RikkiGibson]

This makes it easier to paste baselines in from an AssertEx.Equal failure, since the string values will already be quoted.

I didn't bother to do escaping here, if anything I would like to output a """-string only if the string is detected to contain any characters that need escaping. It's something which can be addressed when we hit it.

[RikkiGibson]

This introduces a problem where every line of a VerifyIL baseline is quoted. I'll have to look at either adjusting that or reverting this change.

[RikkiGibson]

@jcouv @333fred This PR is ready for review. Please take a look at your convenience.

[jcouv]

Done with review pass (iteration 9)

[333fred]

Done review pass

[333fred]

Consider a test like this:

private partial int P1 { get; set; }
partial int P1 { private get => 1; private set {} }

[333fred]

Please test in vs ref readonly with indexer parameters.

[jcouv]

Done with review pass (iteration 16)

[jcouv]

Do we have a similar test (default accessibility vs. explicit accessibility) for accessors?

[RikkiGibson]

I think test ModifierDifference_Accessibility_ExplicitDefault_02 is exercising the scenario adequately because it shows that:

1. explicitly specifying the default accessibility on an accessor is an error
2. a difference in accessor accessibility modifiers between parts is an error

[jcouv]

Done with review pass (iteration 19). Only one test suggestion left

[jcouv]

LGTM Thanks (iteration 19)