fix: stabilize generated ZIP download for large archives

Avoid HTTP/2 streaming failures on large archive downloads by returning the stored ZIP as a buffered HttpResponse with explicit Content-Length.

Also update archive download tests to support both buffered and streaming response objects.

Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>

backend/modules/fleet/views.py

@@ -3063,15 +3063,22 @@ def generated_archive_download(request, archive_id):
         raise DRFValidationError({'detail': 'ZIP arhiva nije dostupna ili je istekla.'})
 
     generated_archive.file.open('rb')
+    try:
+        archive_bytes = generated_archive.file.read()
+    finally:
+        generated_archive.file.close()
+    if not archive_bytes:
+        raise DRFValidationError({'detail': 'ZIP arhiva je prazna ili nedostupna.'})
     filename = generated_archive.filename or _generated_archive_filename_for_user(
         request.user,
         year=generated_archive.year,
         month=generated_archive.month,
         archive_type=generated_archive.archive_type,
     )
-    response = FileResponse(generated_archive.file, content_type='application/zip')
+    response = HttpResponse(archive_bytes, content_type='application/zip')
     response['Content-Disposition'] = f'attachment; filename="{filename}"'
     response['Cache-Control'] = 'private, max-age=3600'
+    response['Content-Length'] = str(len(archive_bytes))
     return response