perf(transcode): eliminate double buffer copy by accepting Bytes instead of &[u8]

Change ImageTranscodePort::get_transcoded signature from &[u8] to Bytes.
- Rayon closure now receives Bytes::clone() (O(1) ref-count) instead of .to_vec() (~5 MB copy)
- Fallback path returns owned Bytes directly instead of Bytes::from(to_vec()) (~5 MB copy)
- Caller passes content.clone() (O(1)) instead of implicit deref

Saves ~10 MB of allocation per transcode call on a 5 MB image.

src/application/ports/transcode_ports.rs

@@ -90,7 +90,7 @@ pub trait ImageTranscodePort: Send + Sync + 'static {
     async fn get_transcoded(
         &self,
         file_id: &str,
-        original_content: &[u8],
+        original_content: Bytes,
         original_mime: &str,
         target_format: OutputFormat,
     ) -> Result<(Bytes, String, bool), DomainError>;

src/application/services/file_retrieval_service.rs

@@ -88,7 +88,7 @@ impl FileRetrievalService {
             return None;
         }
         let format = OutputFormat::WebP;
-        match transcode.get_transcoded(id, content, mime, format).await {
+        match transcode.get_transcoded(id, content.clone(), mime, format).await {
             Ok((transcoded, webp_mime, true)) => {
                 debug!(
                     "🖼️ WebP transcode: {} -> {} bytes ({:.0}% smaller)",