Be careful about getExactClasses of variant types

[MichalStrehovsky]

#123112 started running more outerloop tests, including:

runtime/src/tests/JIT/Methodical/flowgraph/dev10_bug679008/sealedCastVariance.cs

Lines 13 to 32 in 81e28e2

	public static class Repro
	{
	private static bool CheckType(Action<string> a)
	{
	return a.GetType() == typeof(Action<object>);
	}
	[Fact]
	[OuterLoop]
	public static int TestEntryPoint()
	{
	Action<string> a = (Action<object>)Console.WriteLine;
	if (CheckType(a))
	{
	Console.WriteLine("pass");
	return 100;
	}
	Console.WriteLine("FAIL");
	return 101;
	}
	}

We were incorrectly optimizing the a.GetType() == typeof(Action<object>) check to false (type of the parameter is Action<string> and JIT was asking for exact classes of Action<string>; and there are none).

Cc @dotnet/ilc-contrib @EgorBo

[dotnet-policy-service]

Tagging subscribers to this area: @agocke, @dotnet/ilc-contrib
See info in area-owners.md if you want to be subscribed.

[Copilot]

Pull request overview

This PR fixes an incorrect JIT optimization for variant types by adding a variance check to the getExactClasses method. The issue was discovered when outerloop tests started running after PR #123112, specifically the sealedCastVariance.cs test which validates type checks on variant delegate types.

Changes:

• Modified getExactClasses to return -1 (indicating uncertainty) for types with variance, preventing the JIT from making incorrect type assumptions