test: extend 708 with mlet/forrange/mcase shadow rows (#28)

708 pinned param + clet + self-import-skip after #19 landed; the
other 4 sites where check_module_shadow is wired (N_MLET, N_FORRANGE
single, N_FORRANGE tuple, N_MCASE) had no dedicated negative-compile
row. Worker-19's note that all 6 sites use identical call shape is
true today but unpinned by tests; a future asymmetric edit could
silently disable enforcement on one site.

Adds neg_mlet (let-tuple binder shadows), neg_forrange_single
(`for shadowmod of ...`), neg_forrange_tuple (`for (shadowmod, x)
of pairs`), neg_mcase (`case let shadowmod:`). Each errors at the
binder decl line with the canonical
"<kind> '<name>' shadows imported module '<name>'" message.

708's main now runs 8/8 (param, let, mlet, forrange_single,
forrange_tuple, mcase, pos_rename, pos_selfimp).

test/wcc/data/paramshadowmod/neg_forrange_tuple.ww

@@ -0,0 +1,22 @@
+// neg_forrange_tuple — `for (let (shadowmod, x) .. s)` tuple-
+// destructure range loop where the first binder shadows the
+// imported module. N_FORRANGE wires check_module_shadow per-name
+// on the tuple branch (n->list), so the rule fires at the for
+// header even though `x` is innocuous.
+
+use shadowmod;
+
+export fn main() i32 = {
+	let buf: [2]i64;
+	buf[0] = 1i64;
+	buf[1] = 2i64;
+	let s: [](i64, i64);
+	s.ptr = buf.ptr: *(i64, i64);
+	s.len = 1;
+	s.cap = 1;
+	let total: i64 = 0i64;
+	for (let (shadowmod, x) .. s) {
+		total += shadowmod + x;
+	};
+	return total: i32;
+};