fix(parity): init delete_suppressed; gate deleted-path retention

- Initialize PipelineContextSender.delete_suppressed=false: an uninitialized
  true silently skipped the --delete keep-set manifest under -m/--threads,
  so destination extras were never removed.
- Allocate/install the receiver deleted-path observer only when
  report_deletes is set (--info=del / -i / --out-format under --delete), cap
  the retained list at MAX_MANIFEST_ENTRIES, and free already-created lists
  on the receiver-pipeline create failure path.
- Validate report_deletes/report_stats/report_dest_info on receive.
- Correct stale comments (config.h report_deletes, delete_plan.h deleted
  count, multiprocessing.h stats locking, utils.h observer placement).
- Tests: sender delete_suppressed init, report_deletes gating (unit), and
  -m/--delete default delete-after keep-set removal (integration).

src/shared/multiprocessing.h

@@ -93,7 +93,11 @@ typedef struct {
   unsigned long long progress_bytes;
   unsigned long long total_bytes;
   /* Per-type flist / transferred accounting for the rsync --stats breakdown and
-     the progress `to-chk` denominator.  Guarded by mutex_progress. */
+     the progress `to-chk` denominator.  Owned by the sender thread: it is the
+     only writer (the entry/transfer notes in send_chunks_multithreaded) and it
+     reads the totals in its completion tail, so no lock is needed.  This is NOT
+     guarded by mutex_progress (which covers total_files/progress_bytes/
+     total_bytes/sender_done). */
   TransferStats stats;
   bool sender_done;
   atomic_bool cancelled;