session-memory: Phase 4 Measure — baseline, effectiveness, trend (WP-0009)

Closes the loop. metrics.py: fleet metrics (infra-overhead share, error rate,
schema-thrash, token percentiles, success) + persisted baseline trend. effect.py:
before/after per-pattern effectiveness with an improved verdict per metric.
measure entrypoint with trend + --since effectiveness + JSON. Recorded pre-fix
baseline: 27 sessions, overhead median 11.7%, error rate 0.96, schema-thrash 8.
13 new tests; suite 139/139. Capture->Detect->Curate->Distribute->Measure complete.

Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>

tests/test_measure_metrics.py

@@ -0,0 +1,63 @@
+"""Fleet metrics + baseline tests (WP-0009 T01)."""
+
+import os
+import sys
+
+sys.path.insert(0, os.path.dirname(os.path.dirname(os.path.abspath(__file__))))
+
+from session_memory.measure.metrics import (  # noqa: E402
+    aggregate,
+    load_baselines,
+    save_baseline,
+    session_metrics,
+    snapshot,
+)
+
+
+def _digest(tools=None, errors=0, tokens=100, outcome="success"):
+    return {
+        "outcome": outcome,
+        "cost": {"input_tokens": tokens, "output_tokens": 0},
+        "tool_histogram": tools or {"Bash": 10, "Edit": 5},
+        "error_snippets": [{"fingerprint": f"e{i}", "count": 1} for i in range(errors)],
+    }
+
+
+def test_session_metrics_overhead_and_errors():
+    m = session_metrics(_digest(tools={"mcp__state-hub__create_task": 6, "Bash": 4}, errors=2))
+    assert abs(m["infra_overhead_share"] - 0.6) < 1e-9
+    assert m["error_occurrences"] == 2
+    assert m["has_error"] is True
+
+
+def test_aggregate_rates_and_percentiles():
+    digs = [
+        _digest(tools={"mcp__state-hub__x": 8, "Bash": 2}, errors=1, tokens=50),   # 80% overhead
+        _digest(tools={"Bash": 9, "Edit": 1}, errors=0, tokens=200),               # 0% overhead
+        _digest(tools={"ToolSearch": 6, "Bash": 4}, errors=0, tokens=100, outcome="fail"),
+    ]
+    a = aggregate(digs)
+    assert a["n_sessions"] == 3
+    assert a["error_rate"] == round(1 / 3, 3)
+    assert a["success_rate"] == round(2 / 3, 3)
+    assert a["schema_thrash_sessions"] == 1   # the ToolSearch=6 session
+    assert 0 <= a["infra_overhead_share_median"] <= 1
+
+
+def test_aggregate_empty():
+    assert aggregate([]) == {"n_sessions": 0}
+
+
+def test_snapshot_has_timestamp_and_label():
+    s = snapshot([_digest()], label="baseline")
+    assert s["label"] == "baseline"
+    assert "captured_at" in s and s["n_sessions"] == 1
+
+
+def test_baseline_roundtrip_appends(tmp_path):
+    path = str(tmp_path / "baselines.jsonl")
+    save_baseline(snapshot([_digest()], label="a"), path)
+    save_baseline(snapshot([_digest(), _digest()], label="b"), path)
+    rows = load_baselines(path)
+    assert [r["label"] for r in rows] == ["a", "b"]
+    assert rows[1]["n_sessions"] == 2