feat: implement upload pipeline with compression and SSE

Backend:
- POST /api/v1/upload: multipart file upload with caption + hashtags
  - Validates file size against DB config limits (image/video separate)
  - Checks user ban status and event upload lock
  - Saves original to disk under {media_path}/originals/{slug}/
  - Tracks user total_upload_bytes for quota enforcement
  - Extracts hashtags from caption text and explicit CSV field
  - Upserts hashtags and links them to uploads
- PATCH /api/v1/upload/{id}: edit caption and hashtags (owner only)
- DELETE /api/v1/upload/{id}: soft-delete (owner only)
- GET /api/v1/stream: SSE endpoint with 30s keepalive
  - Broadcasts new-upload events to all connected clients
  - Uses tokio broadcast channel for fan-out

Services:
- CompressionWorker: Tokio semaphore-bounded (concurrency=2) background processor
  - Images: resize to 800px wide JPEG preview via image crate
  - PNG originals: lossless compression via oxipng
  - Videos: ffmpeg thumbnail extraction (1 frame at 1s, scaled to 800px)
  - Updates upload record with preview_path/thumbnail_path on completion

Models:
- Upload with full CRUD (create, find, update caption, soft delete, set paths)
- Hashtag with upsert, link/unlink, extract_hashtags() text parser
- UploadDto for API serialization with like/comment counts

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

backend/Cargo.toml

@@ -16,6 +16,8 @@ jsonwebtoken = "9"
 bcrypt = "0.15"
 uuid = { version = "1", features = ["v4", "serde"] }
 chrono = { version = "0.4", features = ["serde"] }
+tokio-stream = { version = "0.1", features = ["sync"] }
+futures = "0.3"
 sha2 = "0.10"
 rand = "0.9"
 anyhow = "1"