fix: container permissions and volume setup

Problems fixed:
- archivum user (UID 1000) could not write to bind-mounted /config or /data
  because Docker (especially on Windows/WSL2) creates volumes owned by root
- /data/wiki and /data/db subdirectories were never created, causing
  "no such file or directory" when the app tried to write files

Changes:
- docker/entrypoint.sh: runs as root, creates /config /data/wiki /data/db,
  chowns them to PUID:PGID, then drops privileges via su-exec
- Dockerfile: install su-exec + git, copy entrypoint, remove build-time
  USER directive (privilege drop happens at runtime via entrypoint)
- docker-compose.yml: switch local dev to named volumes so Docker manages
  ownership automatically (avoids Windows/WSL2 bind-mount permission issues);
  add TZ env var; explicit comments on how to switch to bind mounts

Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>

docker/Dockerfile

@@ -12,8 +12,9 @@ FROM golang:1.22-alpine AS backend-builder
 
 WORKDIR /app/backend
 # Copy go.mod first for layer caching. go.sum is written by go mod download
-# when it doesn't exist yet (skeleton project), or verified when it does.
+# when it doesn't exist yet, or verified when it does.
 COPY backend/go.mod ./
+COPY backend/go.su[m] ./
 RUN go mod download -x
 COPY backend/ ./
 # CGO disabled — pure Go SQLite (modernc.org/sqlite).
@@ -22,23 +23,30 @@ RUN CGO_ENABLED=0 go build -ldflags="-s -w" -o /archivum ./cmd/server
 # ── Stage 3: Runtime image ────────────────────────────────────────────────────
 FROM alpine:3.20
 
-# Allow PUID/PGID override at runtime.
-ARG PUID=1000
-ARG PGID=1000
+# su-exec: lightweight tool to drop privileges (replaces gosu / setpriv).
+# git: needed for wiki commit operations.
+RUN apk add --no-cache su-exec git
 
-RUN addgroup -g "${PGID}" archivum \
- && adduser -u "${PUID}" -G archivum -s /sbin/nologin -D archivum
+# Create the app group and user that the entrypoint will run as.
+# PUID/PGID can be overridden at runtime via environment variables;
+# the entrypoint script uses them rather than these build-time values.
+RUN addgroup -g 1000 archivum \
+ && adduser -u 1000 -G archivum -s /sbin/nologin -D archivum
 
-# Embed frontend into the binary's working directory.
-COPY --from=frontend-builder /app/frontend/dist /srv/archivum/ui
-COPY --from=backend-builder  /archivum            /usr/local/bin/archivum
+COPY --from=frontend-builder /app/frontend/dist    /srv/archivum/ui
+COPY --from=backend-builder  /archivum              /usr/local/bin/archivum
+COPY docker/entrypoint.sh                           /usr/local/bin/entrypoint.sh
+RUN chmod +x /usr/local/bin/entrypoint.sh
 
-# Volumes: config and data live outside the image.
+# Declare the volume mount points. The entrypoint creates subdirectories
+# and fixes ownership before the application starts.
 VOLUME ["/config", "/data"]
 
-ENV DOCKER_PATH=/config
+ENV DOCKER_PATH=/config \
+    UI_DIR=/srv/archivum/ui \
+    PUID=1000 \
+    PGID=1000
 
-USER archivum
+# Entrypoint runs as root so it can chown the volumes, then drops to PUID:PGID.
 EXPOSE 4000
-
-ENTRYPOINT ["/usr/local/bin/archivum"]
+ENTRYPOINT ["/usr/local/bin/entrypoint.sh"]