feat(4636): rule pull-through per surface, and per-rule usage with confirmed lessons on list_rules

retrieval_telemetry gains rule_usage.by_source, the twin of usage.by_source:
distinct rules each surface showed, how many an agent opened, and a ratio
that is null on a bulk delivery. list_rules(with_usage=True, days=N) adds
each rule's usage, windowed, and how many lessons are confirmed instances
of it, so linked and unlinked rules can be compared through MCP.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

tests/test_services_retrieval_telemetry.py

@@ -618,6 +618,54 @@ async def test_the_preload_lands_in_ambient_and_never_in_the_ratio(_dispose_engi
         await cleanup()
 
 
+@pytest.mark.integration
+@pytest.mark.asyncio
+async def test_rule_by_source_says_which_arm_earns_its_lines(_dispose_engine):
+    """The rule ratio blends every arm. Per surface it can say which one is
+    working: distinct rules per arm, the agent opens among them, and no ratio
+    on a bulk delivery."""
+    from grimoire.services.retrieval_telemetry import retrieval_summary
+
+    cleanup = await _rule_events(990016, [
+        # The write arm chose rule 5301 twice — one rule — and 5302 once.
+        (5301, "surfaced", "write_path_rule"),
+        (5301, "surfaced", "write_path_rule"),
+        (5302, "surfaced", "write_path_rule"),
+        # The tool arm chose 5303 and nothing came of it.
+        (5303, "surfaced", "pre_tool_rule"),
+        # The preload carried 5301 as well.
+        (5301, "surfaced", "session_start"),
+        # An agent opened 5301; a person opened 5303, which is not a hint
+        # landing.
+        (5301, "pulled", "mcp_get_rule"),
+        (5303, "pulled", "rest_rule"),
+    ])
+    try:
+        ru = (await retrieval_summary(990016, days=30))["rule_usage"]
+        assert "by_source_failed" not in ru, "the join did not execute"
+        by_source = ru["by_source"]
+
+        wp = by_source["write_path_rule"]
+        assert (wp["rules_surfaced"], wp["rules_pulled"]) == (2, 1)
+        assert wp["pull_through"] == 0.5
+        assert wp["ambient"] is False
+
+        tool = by_source["pre_tool_rule"]
+        assert (tool["rules_surfaced"], tool["rules_pulled"]) == (1, 0)
+        assert tool["pull_through"] == 0.0
+
+        preload = by_source["session_start"]
+        # The pull is counted against every surface that showed the rule,
+        # the upper bound the readout states.
+        assert (preload["rules_surfaced"], preload["rules_pulled"]) == (1, 1)
+        assert preload["pull_through"] is None
+        assert preload["ambient"] is True
+
+        assert "mcp_get_rule" not in by_source, "a pull is not a surface"
+    finally:
+        await cleanup()
+
+
 @pytest.mark.integration
 @pytest.mark.asyncio
 async def test_ambient_alone_reports_no_ratio(_dispose_engine):