Improve handling of wildcard upper bounds in generic method inference

[msridhar]

Test case:

NullAway/nullaway/src/test/java/com/uber/nullaway/jspecify/GenericMethodTests.java

Lines 1061 to 1080 in fed9143

	@Test
	public void genericMethodLambdaArgWildCard() {
	makeHelperWithInferenceFailureWarning()
	.addSourceLines(
	"Test.java",
	"import org.jspecify.annotations.*;",
	"import java.util.function.Function;",
	"@NullMarked",
	"class Test {",
	" static <T, R> R invokeWithReturn(Function <? super T, ? extends @Nullable R> mapper) {",
	" throw new RuntimeException();",
	" }",
	" static void test() {",
	" // legal, should infer R -> Object but then the type of the lambda as ",
	" // Function<Object, @Nullable Object> via wildcard upper bound",
	" Object x = invokeWithReturn(t -> null);",
	" }",
	"}")
	.doTest();
	}

This reflects the suppression in Spring here. Since this comes up often, ideally we would find a way to get rid of this false positive without adding full support for wildcards.