fix: classify source-free aggregate calls correctly (COUNT(*))

[eitsupi]

This PR is based on #12.

Note

This change was written by Claude Code and Codex under my review and direction.
I have verified the diff, run the test suite, and confirmed
the behavior change against a probe crate. Please review it as you would any other
patch. Happy to split this into two PRs if you prefer one fix per PR — the two
commits are independent and can be cherry-picked separately.

Summary

determine_edge_kind already correctly identifies any aggregate function call (COUNT, SUM, MAX, etc.) as EdgeKind::ViaAggregation regardless of its arguments — it only inspects the function name. But that kind is only ever attached to the lineage graph as an edge to an ancestor column:

let kind = determine_edge_kind(expr);
let output = self.graph.add_output(name.clone());
for &anc in &ancestors {
    self.graph.add_edge(anc, output, kind.clone());
}

COUNT(*) has no column ancestor — * inside a function call is FunctionArgExpr::Wildcard, not FunctionArgExpr::Expr, so collect_ancestors never visits it and ancestors is empty. The for loop body never runs, so the correctly-computed kind is silently discarded. Later, derive_transform classifies the output purely from edges that exist in the graph; with zero edges it falls back to TransformKind::Direct — indistinguishable from a genuine literal constant like SELECT 1 AS c.

SELECT COUNT(*) AS c FROM t   -> transform: Direct   (wrong)
SELECT COUNT(x) AS c FROM t   -> transform: Aggregation (correct — has a column ancestor)

Fix

Store the defining expression's intrinsic edge kind on the Output graph node itself (computed at build time, where the AST is visible), and use it as a fallback in derive_transform whenever no ancestor edge exists to classify from. This generalizes beyond COUNT(*) to any zero-ancestor aggregate/conditional/expression, not just this one function.

sources staying empty for COUNT(*) is unchanged and correct — there genuinely is no column source. Only the transform classification changes.

Tests

• Full existing test suite passes unmodified (94 tests across the crate).
• Added select_count_star_is_aggregation_without_sources, modeled on the existing select_aggregate test, asserting COUNT(*) reports transform: Aggregation with empty sources.
• Confirmed the new test fails on the pre-fix code (reverted the implementation, re-ran) and passes with the fix.