Verify semantic model on suppressed expressions

[jcouv]

From discussion with Neal, we think the suppression expression should have matching Type and ConvertedType, which would be the converted type of the underlying expression.
In (Delegate)(() = {}!) and Delegate x = () => {}!;, the suppressed expressions would have type and converted type Delegate.

Other interesting scenarios involving conversions from expressions: null, default, lambdas, method groups, interpolated strings.

Look for uses of BoundNode.Kind and BoundKind.MethodGroup (for instance, GetTypeInfoForNode)

[333fred]

After further discussions with @AlekseyTs, @jcouv, and @gafter, we've decided that the existing Type and ConvertedType fields are correct for suppressed expressions. These examples as written are correct:

Action a = () => {}!; // Type of null, ConvertedType of Action when called on both the inner lambda syntax and when called on the ! syntax
string? s = null;
object o = s!; // Type of string, ConvertedType of object when called on both the inner local reference and when called on the ! syntax.

Currently, the Nullability and ConvertedNullability on these syntax nodes are the same. As a future feature, we would like to have the underlying syntax have the original null state before suppression, but we think that such a feature will be ok to add in the future. This work is tracked by #35412.

[jcouv]

There are references to this issue left.