feat(metadata): SQLite backend, storage reconciliation and local defaults

- pluggable per-domain Repository interfaces; PostgreSQL and SQLite impls
- SQLite on modernc.org/sqlite (WAL, busy_timeout, foreign_keys, immediate tx)
- separate migrations/sqlite and db.MigrateSQLite; store.NewPostgres/NewSQLite
- cmd/reconcile compares object storage with the journal (report, then --delete)
- storage.Backend.List and photos.Repository.AllStorageKeys
- DB_BACKEND/SQLITE_PATH and STORAGE_BACKEND/FS_ROOT selectable; local defaults
  (sqlite ./galo.db, fs ./data)
- Makefile reconcile/reconcile-delete; docs and .env.example updated

backend/internal/storage/storage.go

@@ -8,6 +8,7 @@ import (
 	"net/http"
 	"strconv"
 	"strings"
+	"time"
 
 	"github.com/aws/aws-sdk-go-v2/aws"
 	awsconfig "github.com/aws/aws-sdk-go-v2/config"
@@ -29,6 +30,13 @@ var ErrObjectNotFound = errors.New("object not found")
 // ErrRangeNotSatisfiable is returned when a requested byte range is invalid.
 var ErrRangeNotSatisfiable = errors.New("requested range not satisfiable")
 
+// ObjectInfo describes one stored blob for inventory and reconciliation.
+type ObjectInfo struct {
+	Key     string
+	Size    int64
+	ModTime time.Time
+}
+
 // Backend is the object storage abstraction. Keys are opaque and independent of
 // the backend: every implementation stores and serves the same logical blob.
 // Blobs are always delivered through the app.
@@ -41,6 +49,8 @@ type Backend interface {
 	// ServeObject validates the signed URL and writes the blob, including
 	// Content-Type and range handling, to the response.
 	ServeObject(w http.ResponseWriter, r *http.Request, key string) error
+	// List returns the blobs under prefix for inventory and reconciliation.
+	List(ctx context.Context, prefix string) ([]ObjectInfo, error)
 	Ping(ctx context.Context) error
 	EnsureBucket(ctx context.Context) error
 	// Bucket returns a human-readable identifier for the storage location.
@@ -183,6 +193,36 @@ func (c *S3) URL(_ context.Context, key string) (string, error) {
 	return c.signer.URL(key), nil
 }
 
+// List returns every object under prefix using paginated ListObjectsV2.
+func (c *S3) List(ctx context.Context, prefix string) ([]ObjectInfo, error) {
+	var out []ObjectInfo
+	var token *string
+	for {
+		in := &s3.ListObjectsV2Input{Bucket: aws.String(c.bucket)}
+		if prefix != "" {
+			in.Prefix = aws.String(prefix)
+		}
+		if token != nil {
+			in.ContinuationToken = token
+		}
+		res, err := c.s3.ListObjectsV2(ctx, in)
+		if err != nil {
+			return nil, fmt.Errorf("list objects: %w", err)
+		}
+		for _, o := range res.Contents {
+			out = append(out, ObjectInfo{
+				Key:     aws.ToString(o.Key),
+				Size:    aws.ToInt64(o.Size),
+				ModTime: aws.ToTime(o.LastModified),
+			})
+		}
+		if res.IsTruncated == nil || !*res.IsTruncated {
+			return out, nil
+		}
+		token = res.NextContinuationToken
+	}
+}
+
 // ServeObject validates the signed URL and streams the blob from S3. It passes
 // the client's Range header through to S3, so byte ranges are handled natively,
 // and answers HEAD via HeadObject. It writes the response itself.