wcc: converge let-IDENT memcpy, IDENT-assign recv, nested struct call-recv onto structabisize

Three more wwstage cgen sites still used unrounded structnaturalsize where
cstage rounds via lu->size — pre-existing gate-blind cs!=ww latents the #169b
reviewer surfaced: cgenstmt N_LET struct-IDENT memcpy (let p2: T = p1; twin
cgen.c:7869), cgenexpr N_ASSIGN N_IDENT-lhs register RECV (s = mk(); twin
cgen.c:4700-4737), and cgenutil's nested struct N_CALL recv inside
cgstructlitfill (twin cgen.c:2121).

Converge all three onto structabisize, completing the same-class closure
started by #169 and continued by #169b. Also corrected the inline comment at
cgenutil.ww:3273-3286 that wrongly claimed fl->type->size was natural
(check.c:760 sets ABI). sretretsize at cgenutil.ww:1301 is gate-equivalent
natural and is left alone.

Probe 698 +3 rows (one per converged site) with cs==ww .s byte-cmp and a
pre-fix-rebuild discriminator. 990-997 byte-id hold.

test/wcc/698_cgreturn_struct.c

@@ -242,6 +242,52 @@ static const struct row rows[] = {
 	  "    return z.a + z.b + z.c;\n"
 	  "};\n",
 	  0 },
+	/* `let p2: T = p1;` ident memcpy. struct{i64,i32} natural 12,
+	 * maxalign 8 → cstage N_LET sets sz = lu->size = 16 (cgen.c:7533,
+	 * struct-IDENT branch :7869) and emits MOVQ+MOVQ for the per-qword
+	 * copy. Pre-fix wwstage sized the run on structnaturalsize (12) →
+	 * MOVQ+MOVL. The byte-id catches the 4B widen; both stages
+	 * round-trip the value correctly because the high 4B of the rhs
+	 * slot were zero-init'd by the source let. */
+	{ "let_ident_memcpy_pair_i64_i32",
+	  "type pt = struct { a: i64, b: i32 };\n"
+	  "fn main() i32 = {\n"
+	  "    let p1: pt = pt { a = 7i64, b = 9i32 };\n"
+	  "    let p2: pt = p1;\n"
+	  "    return p2.b - 9;\n"
+	  "};\n",
+	  0 },
+	/* `s = mk();` ident-lhs register RECV (sibling of dot_recv_call_*).
+	 * struct{i64,i32} natural 12, maxalign 8 → cstage N_ASSIGN-IDENT
+	 * sets sz = lu->size = 16 (cgen.c:4704) and stores AX/DX with
+	 * MOVQ+MOVQ. Pre-fix wwstage used structnaturalsize (12) →
+	 * MOVQ+MOVL. byte-id discriminator regardless of value readback. */
+	{ "assign_ident_call_recv_pair_i64_i32",
+	  "type pt = struct { a: i64, b: i32 };\n"
+	  "fn mk() pt = { return pt { a = 7i64, b = 9i32 }; };\n"
+	  "fn main() i32 = {\n"
+	  "    let s: pt = pt { a = 0i64, b = 0i32 };\n"
+	  "    s = mk();\n"
+	  "    return s.b - 9;\n"
+	  "};\n",
+	  0 },
+	/* Nested struct N_CALL inside a structlit field: the outer literal
+	 * has field `v` whose type is struct{i64,i32}; the value is mk().
+	 * cstage's cg_structlit_fill reads fl->type->size at the nested-
+	 * call branch (cgen.c:2121) — per check.c:760 that's the maxalign-
+	 * rounded ABI extent (16). Pre-fix wwstage cgstructlitfill used
+	 * structnaturalsize (12) → MOVQ+MOVL where cstage emits MOVQ+MOVQ.
+	 * mk returns via a structlit so the producer side stays out of the
+	 * comparison; only the nested-call-recv ladder shifts. */
+	{ "structlit_nested_call_recv_pair_i64_i32",
+	  "type pt = struct { a: i64, b: i32 };\n"
+	  "type box = struct { v: pt };\n"
+	  "fn mk() pt = { return pt { a = 7i64, b = 9i32 }; };\n"
+	  "fn main() i32 = {\n"
+	  "    let b: box = box { v = mk() };\n"
+	  "    return b.v.b - 9;\n"
+	  "};\n",
+	  0 },
 };
 
 static int