audit: lock DTS rate table, fix sniff overflow-scan, cover decrypt loss

Round-10 findings from the 10-phase release audit:

- A finder claimed the DTS SFREQ→rate table was wrong at 11/12; verified
  it against ffmpeg's avpriv_dca_sample_rates (12k/24k/48k/96k/192k at
  11-15) — the table is CORRECT. Added a test that locks the full table so
  it can't be mis-"fixed".
- sniff_video_codec advanced 3 bytes after a matched start code, re-reading
  the code byte as an overlapping start code; skip the full 4-byte marker.
- Guard the HD-DVD next_id title counter with saturating_add so a crafted
  disc with >65536 clips can't overflow (panic in debug).
- Add a test that an undecryptable unit (DecryptFailed) is zero-filled and
  counted as loss through ExtractResult (complete=false, bytes_lost>0) —
  the recovery-seam consolidation folded that bucket into bytes_unreadable.

src/disc/hddvd.rs

@@ -152,7 +152,9 @@ fn sniff_video_codec(es: &[u8]) -> Option<Codec> {
                 _ if (code & 0x9F) == 0x07 => saw_h264_sps = true,
                 _ => {}
             }
-            i += 3;
+            // Skip the whole consumed `00 00 01 <code>` marker (4 bytes) so the
+            // code byte isn't re-read as the start of an overlapping start code.
+            i += 4;
         } else {
             i += 1;
         }
@@ -423,7 +425,7 @@ impl Disc {
                 content_format: ContentFormat::MpegPs,
                 codec_privates: Vec::new(),
             });
-            next_id += 1;
+            next_id = next_id.saturating_add(1);
         }
 
         // Every remaining clip is its own title (unchanged behaviour). Iterated in
@@ -461,7 +463,7 @@ impl Disc {
                 content_format: ContentFormat::MpegPs,
                 codec_privates: Vec::new(),
             });
-            next_id += 1;
+            next_id = next_id.saturating_add(1);
         }
         titles
     }