quak backup writes each file's ML data into its JSON (closes #163)

lib.backup() now waits for an ML data fetch, joining the one its refresh
started or starting one, before it writes the per-file JSON, and each
file's JSON carries the cached payload as mlData. When the fetch fails,
each file with no cached ML data gets mlDataError and an entry in
failures.json, so the result counts it as failed and quak backup exits 1;
the next run fetches again.

Judgement call: the wait comes after the originals are downloaded, so the fetch runs alongside the downloads.
Judgement call: a failed ML fetch is recorded per file in failures.json, which is how the exit code goes non-zero without changing src/cli-commands.ts.

Model: opus-5-5

TODO.md

@@ -25,6 +25,13 @@ declares one.
 
 # Completed Steps
 
+- 2026-10-05: `quak backup` writes each file's ML data (its faces and CLIP
+  embedding) into the file's JSON as `mlData`, the payload
+  `lib.mldata.forFile()` returns (issue 163). `lib.backup()` waits for an ML
+  data fetch before it writes the JSON files. When that fetch fails, each file
+  whose ML data is not cached gets the reason in `mlDataError` and counts as
+  failed, and the next run fetches it again.
+
 - 2026-10-03: The download-albums example test no longer fails on a disk with
   under 50 GiB free (issue 160). It opens its libraries with
   `freeBelowBytes: 0`, so the free space of the disk it runs on cannot shrink