test(java): pin enum constants surviving alongside enum methods

[DeusData]

Closes the gap that #984 by @sahil-mangla exposed. Credited via Co-authored-by.

The feature already works

Java enum methods resolve correctly on main today: find_class_member_body descends into enum_body_declarations (Java-gated) while find_class_body keeps returning enum_body, which is what extract_enum_members needs to reach the constants — they are siblings of enum_body_declarations, not children. I verified that empirically with a probe asserting the enum label, its methods and its constants together; it passes unmodified.

But the distinction was unguarded

java_enum_dedup_preserves_calls_issue1234 asserted the methods and the absence of duplicate Function defs — but never asserted the constants. So collapsing those two lookups into one would have passed the entire suite while silently dropping every enum constant from the graph.

That is exactly what #984 proposed: redirecting the shared find_class_body. Its own Method assertion would have gone green, and the loss would have shipped invisibly.

The guard

Two assertions added to the existing test rather than a near-duplicate test, since the fixture already has the right shape.

Verified binding: pointing extract_enum_members at find_class_member_body — the collapse #984 proposed — reddens exactly the new Variable assertions and nothing else.

The comment at the assertions names why the two lookups differ, so the next person to touch enum extraction finds the reasoning rather than rediscovering it through a silent regression.