fix(core,fuse): un-park abandoned creates, retire orphan staged bytes, serialise refreshes

Four faults found by reading the journal for issues that were never fixed,
tracked as B94-B97 in docs/BUGS.md.

A create parked behind a temp-file rename (B70) had no exit but that rename,
so 43 uploads sat in the queue for weeks holding their staged bytes. The
primary drain worker now sweeps parked creates while the queue is idle: one
whose node is gone is discarded, and one older than PARK_EXPIRY_MS that nobody
holds open is un-parked rather than dropped, since staged bytes are the only
copy of the write. An open node stays parked however old it is.

.tmp and .temp leave TRANSIENT_SUFFIXES. The list served two mechanisms with
different consequences: ignoring a name during mirror reconcile is reversible,
parking a create is not, and a Takeout index.tmp is a final name that nothing
will ever rename.

Staged blobs that match no queued op and name no node were kept forever and
warned about once per blob per mount (95 blobs, ~445 MiB, oldest two months).
Blobs past STAGING_ORPHAN_RETAIN are retired; the rest are one aggregated
warning.

Refresh control requests now run one per scope. Every scope is individually
cheap, but duplicates contend for the database lock the running refresh holds,
which left five in flight for twenty-three minutes. request_kind names the
scope so the report says which view wedged. The photo metadata pass also stops
at shutdown instead of running into the cancelled runtime and logging one
"skipping undecryptable photo" per node.

docs/CHANGELOG.md

@@ -9,6 +9,33 @@ release ships. Migrations are forward-only — a database written by a newer cli
 refuse-to-open, not a downgrade, so rolling back a release means restoring the cache from
 scratch (user data in `staging/` and `recovery/` is never touched by this).
 
+## [1.11.1] — 2026-09-19
+
+Schema: **30** (unchanged).
+
+### Fixed
+- **An upload parked behind a temp-file rename no longer stays parked forever.** A create for a
+  name that looks like a writer's temporary file waits for the rename to the final name. If that
+  rename never came — the writer crashed, the file was deleted, or the name was never temporary —
+  the upload sat in the queue permanently, holding its staged bytes. The queue now sweeps parked
+  creates while it is idle: one whose file is gone is dropped, and one that has been parked for
+  over an hour with nobody holding it open is uploaded under the name it has. An open file is
+  left alone however old it is (B94).
+- **A file whose real name ends in `.tmp` uploads normally.** `.tmp` and `.temp` were treated as
+  writer temp-file suffixes, so ordinary files — a Google Photos Takeout `index.tmp`, for
+  instance — were parked waiting for a rename that was never going to happen (B95).
+- **Staged bytes that belong to nothing are cleaned up.** A staged blob that matches no queued
+  upload and names no node used to be kept forever and warned about once per blob on every mount.
+  Blobs older than 30 days are now retired with their size logged; younger ones are kept, but
+  reported as a single line instead of one per blob (B96).
+- **Refreshing the gallery repeatedly no longer wedges the control socket.** Each refresh scope
+  now runs one at a time; asking again while one is running is answered immediately rather than
+  repeating work that contends with it for the database. `pdfs diagnostics` also names the scope
+  of a running refresh (B97).
+- **Shutting down during a photo refresh no longer logs a wall of "undecryptable photo"
+  warnings.** The metadata pass stops when teardown begins instead of running into the cancelled
+  runtime. Nothing was wrong with those photos; the next refresh reads them again (B97).
+
 ## [1.11.0] — 2026-09-19
 
 Schema: **30** (`photos.content_hash` / `main_uid` / `group_key`); SDK bumped to `proton-sdk` /