Cycle 5 Phase 3 closed: M1 PASS via bench pointer-convention fix

The previous "layout mismatch" deferral was a one-line bench bug:
NEON expects the caller to pass tmp pointing at the 8x8 block
origin (after the 2*16+2 padding skip), but the bench passed the
raw padded-buffer origin. C ref does the advance internally, so it
filtered the correct block; NEON filtered a (+2 rows, +2 cols)
shifted region. Diagonal-shift trace in the partial doc was
exactly that.

Fix: tmps + i*TMP_INTS + (2*TMP_W + 2) for NEON calls.

Results:
  M1: 10000/10000 bit-exact (100.0000%), all 8 dirs balanced
  M3: 3.809 Mblock/s (consistent with 3.923 from longer window)

Phase 4 unblocked; predicted R5 = 0.02-0.05 (deep RED) per earlier
analysis. Will build QPU CDEF anyway for cycle-completeness +
V4L2 dispatch-path existence.

- tests/bench_neon_cdef.c: 3-line tmp pointer fix
- docs/k5_cdef_phase3.md: supersedes k5_cdef_phase3_partial.md

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>

tests/bench_neon_cdef.c

@@ -113,11 +113,16 @@ static int correctness_check(uint64_t seed, int n)
         tmp_center_to_dst(dst_a, tmp);
         memcpy(dst_b, dst_a, DST_BYTES);
 
+        /* C ref advances tmp internally by +2*stride+2.
+         * NEON expects the caller to pass the already-advanced pointer
+         * (i.e. pointer to the block-data origin, not the padded-buffer
+         * origin). Hence the tmp+34 for the NEON call. */
         daedalus_cdef_filter_8x8_pri_sec_ref(
             dst_a, DST_W, tmp, pri, sec, dir, damping, 8);
         dav1d_cdef_filter8_8bpc_neon(
-            dst_b, DST_W, tmp, pri, sec, dir, damping, 8,
-            /* edges = */ 0);   /* != 0xf → non-edged path, uint16 tmp w/stride 12 */
+            dst_b, DST_W, tmp + (2 * TMP_W + 2),
+            pri, sec, dir, damping, 8,
+            /* edges = */ 0);   /* uint16 tmp non-edged path */
 
         if (memcmp(dst_a, dst_b, DST_BYTES) != 0) {
             if (mismatches < 3) {
@@ -180,7 +185,7 @@ static void throughput_neon(uint64_t seed, int n_blocks, double duration_s)
     for (int i = 0; i < n_blocks; i++)
         dav1d_cdef_filter8_8bpc_neon(
             work_dst + (size_t)i * DST_BYTES, DST_W,
-            tmps + (size_t)i * TMP_INTS,
+            tmps + (size_t)i * TMP_INTS + (2 * TMP_W + 2),
             pris[i], secs[i], dirs[i], damps[i], 8, 0);
 
     double t0 = now_seconds();
@@ -191,7 +196,7 @@ static void throughput_neon(uint64_t seed, int n_blocks, double duration_s)
         for (int i = 0; i < n_blocks; i++)
             dav1d_cdef_filter8_8bpc_neon(
                 work_dst + (size_t)i * DST_BYTES, DST_W,
-                tmps + (size_t)i * TMP_INTS,
+                tmps + (size_t)i * TMP_INTS + (2 * TMP_W + 2),
                 pris[i], secs[i], dirs[i], damps[i], 8, 0);
         done += n_blocks;
     }