L4 port lattice is partially disconnected from the statement ddg (3 of 4 binding-edge classes missing)

[rahlk]

Sibling finding from codellm-devkit/codeanalyzer-python#115 (propagation sweep; fixed there for 1.1.1).

src/schema/v2/dataflow.ts emits only ONE of the four statement ↔ port ddg binding classes: the return/global → @formal_out routing (prov:["reaching-defs"]). The other three are not emitted:

• def stmt → actual_in:k (argument binding at the call site — emitL4 creates the actual_in vertex and its param_in edge, but no ddg edge feeds the port)
• actual_out → use stmt (return-value binding back into the caller)
• formal_in:k → first-use stmts (parameter flow inside the callee — currently carried only by @entry → stmt edges, so entering via param_in dead-ends at formal_in)

Consequence, same as the python issue: an end-to-end flows_to(def_stmt, callee_formal) or return-value walk cannot cross a call except through the formal_out leg. The python fix (codeanalyzer-python PR for #115) emits the full set with prov:["reaching-defs"] — the label this repo already uses — so the vocabulary is shared; this is emission-only.

Suggested acceptance: on a 2-function sample (y = build(x)), assert ddg contains def→actual_in, actual_out→callsite, formal_in→first-use, return→formal_out, all reaching-defs-tagged, endpoints present in body.