fix: downloads stop leaving files on disk that the library doesn't use

- An archive whose images were extracted was kept at its download path
  as well as in the attachments store. Both download paths now remove
  the downloaded copy.
- A downloaded file the import filters reject (too small, transparent,
  single colour, unreadable) was left on disk with no record. A later
  walk found it, imported it again and rejected it again, every time.
  Rejected files are now removed; the seen ledger keeps them from being
  downloaded again.
- Both paths share one policy, Importer.discard_after_attach, which
  never removes a file an ImageRecord points at.
- An external-link fetch that fails, or whose worker was killed, left
  partial files (megatools' .megatmp.*) in the link's dir, and a killed
  fetch left its link 'downloading' for good. A fetch now clears
  unrecorded leftovers before starting and after failing. The sweep
  returns a link to retry once it has been 'downloading' past the
  task's hard limit, using a new external_link.started_at column
  (migration 0117).
- The orphaned temp-file sweep now recognises .megatmp.* as well as
  .part and .partial.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

backend/app/services/importer.py

@@ -1401,6 +1401,47 @@ class Importer:
         self.session.commit()
         return ImportResult(status="imported", image_id=record.id)
 
+    def discard_after_attach(self, path: Path, result: ImportResult) -> bool:
+        """Delete a downloaded file that `attach_in_place` has finished with,
+        when keeping it would only leave a second copy or a rejected file on
+        disk. Returns True when the file was removed. Both download paths (a
+        subscription walk's import step and the external-link worker) apply
+        this one policy:
+
+        - attached: the attachments store holds the canonical copy.
+        - imported or superseded ARCHIVE: its new images are in the library and
+          the archive itself is in the attachments store.
+        - duplicate skip: the library already holds the content.
+        - soft skip (too small, too transparent, single colour, not a readable
+          image): the import filters rejected it. A file left on disk with no
+          record is picked up by the next walk and imported again; the seen
+          ledger is what stops it being downloaded again.
+        - failed: the archive probe crashed, and a retry would crash again.
+
+        An imported or superseded media file is kept: it IS the record's file.
+        So is any path an ImageRecord points at, whatever the status, because a
+        re-fetch can write identical bytes over a record's own file."""
+        if result.status in ("imported", "superseded"):
+            redundant = is_archive(path)
+        else:
+            redundant = result.status in ("attached", "skipped", "failed")
+        if not redundant:
+            return False
+        owned = self.session.execute(
+            select(ImageRecord.id).where(
+                ImageRecord.path.in_({str(path), str(path.resolve())})
+            ).limit(1)
+        ).first()
+        if owned is not None:
+            log.info("keeping %s: it is image %s's file", path.name, owned.id)
+            return False
+        try:
+            path.unlink(missing_ok=True)
+        except OSError as exc:
+            log.warning("could not remove %s: %s", path, exc)
+            return False
+        return True
+
     def _upsert_artist(self, name: str) -> Artist:
         slug = slugify(name)
         artist = self.session.execute(