wcc: kind-filter type-position name resolution so a value can't shadow a same-named type (#225)

resolve_typename used the kind-blind scope_lookup_prefer, so a same-named value binding (param/let) in a closer scope hid the type it shadowed, wrongly rejecting valid Hare like 'fn f(off: off)'. wwstage already separates type/value namespaces; this aligns the cstage frontend up. New scope_lookup_type skips non-SK_TYPE syms and keeps scanning, preserving same-module preference. Byte-id-neutral: the new branch fires only on the old 'unknown type' error path.

Makefile

@@ -335,6 +335,7 @@ TESTS	= $(BIN)/test_smoke $(BIN)/test_lex $(BIN)/test_parse $(BIN)/test_check \
 	  $(BIN)/test_fmt_vstream_compositions_run \
 	  $(BIN)/test_fieldfn_leaf_collide_run \
 	  $(BIN)/test_amp_fn_assign_run \
+	  $(BIN)/test_type_value_shadow_run \
 	  $(BIN)/test_xmod_alias_struct_collide_run \
 	  $(BIN)/test_xmod_variant_match \
 	  $(BIN)/test_named_ptr_alias_variant_widen \
@@ -735,6 +736,16 @@ $(BIN)/test_amp_fn_assign_run: test/wcc/783_amp_fn_assign_run.c \
 		$(LIB)/libwwrt.a | $(BIN)
 	$(CC) $(CFLAGS) -o $@ $<
 
+# #225: a value binding (param/let/fn) that shadows a same-named TYPE
+# must not hide it in type-annotation or cast position. cstage was
+# kind-blind in resolve_typename; wwstage already separated the type and
+# value namespaces. Driver build + run on both stages + cs.s == ww.s.
+$(BIN)/test_type_value_shadow_run: test/wcc/788_type_value_shadow_run.c \
+		$(BIN)/ww $(BIN)/w6c $(BIN)/w6a $(BIN)/w6l \
+		$(BIN)/ww_ww $(BIN)/w6c_ww $(BIN)/w6a_ww $(BIN)/w6l_ww \
+		$(LIB)/libwwrt.a | $(BIN)
+	$(CC) $(CFLAGS) -o $@ $<
+
 # #223: a struct-field access through a pointer-ALIAS receiver whose
 # leaf collides with another module's same-leaf STRUCT. cstage driver
 # build + run for runtime, raw w6c vs w6c_ww .s cmp on the combined.ww