feat(model-grounded-detection): tasks 2.1-2.2, the Joern seam and the taint verdict

The whole engine integration is one subprocess boundary: joern-parse builds a
cpg.bin, joern --script runs a shipped CPGQL query that prints its JSON between
fence markers, because Joern's stdout carries a banner and a REPL prompt around
whatever a script emits. Nothing is imported in-process, so the core install
keeps dependencies = [] -- a test walks the seam's AST to assert it imports
only the standard library and names no JVM binding.

Fail closed everywhere. A build or query that errors, times out, or prints
something unparseable yields no verdict and the caller reports suspicion. The
taint verdict distinguishes this from absence deliberately: a query returning
None means "the engine could not decide", which is not "there is no flow", and
conflating them is how a missing engine silently becomes a clean bill of health.

The verdict is where the ladder earns its keep. An unsanitized source->sink
path is model_entailed -- the graph establishes it and the LLM's claim is no
longer load-bearing. A path a sanitizer may break is model_corroborated: the
flow is real, but whether the sanitizer suffices is the semantic residual the
LLM answers, checked against this verdict rather than averaged over runs. The
chosen flow is picked by a total order (shortest, then lexicographic), so two
runs over one CPG cannot disagree -- Req 6.4 asserted directly.

The query itself is the part our flat IR could not do: reachableByFlows walks
the PDG, so `q = "..." + request.args["x"]; execute(q)` is found, which is
exactly the one-hop limit that held measured entailment at 2.2%.

Capability detection reuses semantic/engines.joern_available rather than
growing a second notion of "installed". The audit manifest is regenerated; the
new modules are named by Req 4/5/6 rather than counted as orphans.

Task 2.3, the go/no-go measurement, needs Joern installed and has not run.

Co-Authored-By: Claude Opus 5 (1M context) <[email protected]>
Claude-Session: https://claude.ai/code/session_01T7XD7x5KonMdmGRkp63EQG

src/openultrasast/model/ladder.py

@@ -0,0 +1,48 @@
+"""The evidence ladder: what established a finding (model-grounded-detection Req 5).
+
+Every finding carries the rung that established it, so a deterministic model result is never confused with an
+LLM suspicion. The two middle rungs are the point of this feature:
+
+    suspicion           a site was enumerated and something flagged it
+    model_corroborated  the LLM's claim is consistent with the CPG -- the source it names reaches the sink it
+                        names, or a guard its siblings carry is absent
+    model_entailed      the CPG establishes it alone, independent of trusting the LLM
+    execution_confirmed a sandbox reproduced it (deferred tier, memory/C first)
+
+`model_corroborated` and above are established with no model call, and deterministically: run the checker
+twice over the same CPG and the verdict sequence is identical. That is what removes the run-to-run noise the
+prior architecture tried to average away.
+"""
+
+from __future__ import annotations
+
+from dataclasses import dataclass
+from enum import StrEnum
+
+
+class Rung(StrEnum):
+    SUSPICION = "suspicion"
+    CORROBORATED = "model_corroborated"
+    ENTAILED = "model_entailed"
+    EXECUTION_CONFIRMED = "execution_confirmed"
+
+
+# Ascending strength; `_ORDER.index` is the comparison, so a new rung cannot be silently mis-ranked.
+_ORDER = (Rung.SUSPICION, Rung.CORROBORATED, Rung.ENTAILED, Rung.EXECUTION_CONFIRMED)
+
+
+def at_or_above(rung: Rung, threshold: Rung) -> bool:
+    return _ORDER.index(rung) >= _ORDER.index(threshold)
+
+
+@dataclass(frozen=True)
+class Verdict:
+    """What the model established, and the evidence it established it from."""
+
+    rung: Rung
+    family: str
+    witness: str = ""  # the taint path or the dominance fact; empty only at `suspicion`
+    contradiction: str = ""  # why an LLM claim was dropped, when it was
+
+
+__all__ = ["Rung", "Verdict", "at_or_above"]