Check grid and room output on exact shapes and report shorts

The grid and room engines write copper they never check. After a grid or
room run the result is now put through the exact engine's geometry (its
pads, fixed copper and keepouts; wires as boxes or slanted slabs; vias as
their pads) and the report gives the shorts and clearance violations, with
examples, instead of a blanket warning. --strict exits 3 when there are any;
by default the exit status stays 0 so the JAR drop-in contract holds.

The exact engine's static-copper setup moves into insert_static_copper so
both use it; exact runs are byte-identical before and after.

tests/test_cli.py

@@ -82,9 +82,23 @@ def test_cli_default_engine_is_exact(tmp_path):
 
 
 @pytest.mark.parametrize("engine", ["grid", "room"])
-def test_unchecked_engines_say_so(tmp_path, capsys, engine):
+def test_unchecked_engines_are_checked_after_routing(tmp_path, capsys, engine):
     assert main([str(FIXTURE), "--engine", engine, "-do", str(tmp_path / "out.ses")]) == 0
-    assert "does not check its copper" in capsys.readouterr().err
+    assert f"{engine} output checked on exact shapes: 0 short(s), 0 clearance" in capsys.readouterr().err
+
+
+def test_strict_fails_a_run_whose_copper_is_not_clean(tmp_path, monkeypatch, capsys):
+    from rpmroute import cli
+    from rpmroute.route.verify import VerifyReport, Violation
+
+    bad = VerifyReport(shorts=1, short_examples=[Violation("A", "B", "F.Cu", 1.0, 2.0)])
+    monkeypatch.setattr(cli, "verify_result", lambda dsn, result: bad)
+    out = tmp_path / "out.ses"
+    assert main([str(FIXTURE), "--engine", "grid", "-do", str(out)]) == 0
+    assert main([str(FIXTURE), "--engine", "grid", "--strict", "-do", str(out)]) == 3
+    assert out.exists()  # the session is still written
+    err = capsys.readouterr().err
+    assert "short: A x B on F.Cu at (1.00, 2.00) mm" in err and "not DRC-clean" in err
 
 
 def test_the_exact_engine_does_not_warn(tmp_path, capsys):