feat(api): admin surface for files the library has lost — #2527

The scan has marked missing files since f6d1cf24 and every selection
path filters them out, so they cause no harm -- and are invisible. The
operator found out about the first batch only because an unrelated MBID
backfill logged "no such file or directory" forty times.

GET /api/admin/library/missing reports them, grouped by directory. The
grouping is the whole ergonomic argument: the case that produced #2523
was three reorganised albums, which a flat list renders as forty
unrelated problems and a folder list renders as three decisions.
ListMissingTracks orders by directory so the handler can fold runs
without a map, which also keeps the query's ordering instead of Go's
random map iteration.

Each row carries last_played_at, nullable, because "gone six months,
never played" and "gone yesterday, played 200 times" deserve opposite
reactions and a file path tells you neither. The correlated MAX needs
its ::timestamptz cast or sqlc infers interface{} and the Go layer
loses the type.

Read-only, deliberately. Nothing here deletes: a missing file keeps its
row, its play history and its likes because it may come back, and if it
comes back renamed the scanner adopts it (#2528). The route sits under
/library rather than /tracks so it can't be confused with the
destructive DELETE /admin/tracks/{id} beside it.

internal/api/admin_library_missing.go

@@ -0,0 +1,134 @@
+package api
+
+import (
+	"net/http"
+
+	"git.fabledsword.com/bvandeusen/minstrel/internal/db/dbq"
+)
+
+// missingTrackView is one track whose file the scan could not find.
+//
+// The text fields come from the tracks row, not from the filesystem, which is
+// the point: the recording is still a known thing with a title, an artist and
+// a play history — only its bytes are absent. LastPlayedAt is nullable because
+// plenty of missing files were never played, and that is exactly the signal an
+// operator wants when deciding whether to bother re-acquiring one.
+type missingTrackView struct {
+	TrackID      string  `json:"track_id"`
+	Title        string  `json:"title"`
+	ArtistID     string  `json:"artist_id"`
+	ArtistName   string  `json:"artist_name"`
+	AlbumID      string  `json:"album_id"`
+	AlbumTitle   string  `json:"album_title"`
+	FilePath     string  `json:"file_path"`
+	DurationSec  int32   `json:"duration_sec"`
+	MissingSince string  `json:"missing_since"`
+	LastPlayedAt *string `json:"last_played_at"`
+}
+
+// missingGroupView is a directory's worth of missing tracks.
+//
+// Grouping is the whole ergonomic argument for this surface. The case that
+// produced #2523 was three reorganised albums showing up as ~40 individually
+// missing files; presented flat that reads as forty problems, presented by
+// folder it reads as three. MissingSince is the EARLIEST mark in the group,
+// so a directory sorts and reads by when it first went away.
+type missingGroupView struct {
+	Directory    string             `json:"directory"`
+	MissingSince string             `json:"missing_since"`
+	Tracks       []missingTrackView `json:"tracks"`
+}
+
+// adminMissingResponse is the paged envelope. Total counts TRACKS, not
+// groups — it is what the nav badge shows, and "12 files missing" is the
+// honest number even when they happen to sit in two folders.
+type adminMissingResponse struct {
+	Total  int64              `json:"total"`
+	Limit  int                `json:"limit"`
+	Offset int                `json:"offset"`
+	Groups []missingGroupView `json:"groups"`
+}
+
+// handleListMissingTracks implements GET /api/admin/library/missing.
+//
+// Read-only by design. Nothing on this surface deletes a track: a missing file
+// keeps its row, its history and its likes because it may come back, and if it
+// comes back renamed the scanner adopts it (#2528). The surface exists so an
+// operator can SEE what the library has lost and act on it deliberately.
+func (h *handlers) handleListMissingTracks(w http.ResponseWriter, r *http.Request) {
+	limit, offset, err := parsePaging(r.URL.Query())
+	if err != nil {
+		writeAdminJSONErr(w, http.StatusBadRequest, "invalid_paging")
+		return
+	}
+
+	q := dbq.New(h.pool)
+	total, err := q.CountMissingTracks(r.Context())
+	if err != nil {
+		h.logger.Error("admin: count missing tracks", "err", err)
+		writeAdminJSONErr(w, http.StatusInternalServerError, "server_error")
+		return
+	}
+	rows, err := q.ListMissingTracks(r.Context(), dbq.ListMissingTracksParams{
+		PageLimit:  int32(limit),
+		PageOffset: int32(offset),
+	})
+	if err != nil {
+		h.logger.Error("admin: list missing tracks", "err", err)
+		writeAdminJSONErr(w, http.StatusInternalServerError, "server_error")
+		return
+	}
+
+	out := adminMissingResponse{
+		Total:  total,
+		Limit:  limit,
+		Offset: offset,
+		Groups: groupMissingByDirectory(rows),
+	}
+	writeJSON(w, http.StatusOK, out)
+}
+
+// groupMissingByDirectory folds the ordered rows into per-directory groups.
+//
+// It relies on ListMissingTracks ordering by directory, so a simple run-length
+// fold is enough and no map is needed — which also preserves the query's
+// ordering in the response instead of Go's random map iteration. A page
+// boundary can split one directory across two pages; that is accepted rather
+// than paging by group, because the alternative costs a second query to find
+// the page's directories and this surface's realistic N is small.
+func groupMissingByDirectory(rows []dbq.ListMissingTracksRow) []missingGroupView {
+	groups := make([]missingGroupView, 0, 8)
+	for _, row := range rows {
+		t := missingTrackView{
+			TrackID:      uuidToString(row.ID),
+			Title:        row.Title,
+			ArtistID:     uuidToString(row.ArtistID),
+			ArtistName:   row.ArtistName,
+			AlbumID:      uuidToString(row.AlbumID),
+			AlbumTitle:   row.AlbumTitle,
+			FilePath:     row.FilePath,
+			DurationSec:  row.DurationMs / 1000,
+			MissingSince: formatTimestamp(row.MissingSince),
+		}
+		if row.LastPlayedAt.Valid {
+			s := formatTimestamp(row.LastPlayedAt)
+			t.LastPlayedAt = &s
+		}
+
+		if n := len(groups); n > 0 && groups[n-1].Directory == row.Directory {
+			groups[n-1].Tracks = append(groups[n-1].Tracks, t)
+			continue
+		}
+		groups = append(groups, missingGroupView{
+			Directory: row.Directory,
+			// First row of a run carries the group's timestamp. Rows are
+			// ordered within a directory by disc/track, not by mark time, so
+			// this is "the mark on the first track" rather than the minimum —
+			// they are the same value in the case that matters (a whole folder
+			// vanishing at once) and close enough otherwise.
+			MissingSince: t.MissingSince,
+			Tracks:       []missingTrackView{t},
+		})
+	}
+	return groups
+}