fix(models): correct DFlash2 Q8_0 drafter download URL

The URL used the filename as the Hugging Face repo name. Also align
the compose fallback for LLAMA_DRAFT_MODEL_FILE with Q8_0.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

docker-compose.yml

@@ -29,7 +29,7 @@ services:
       --cache-type-v q8_0
       --cache-ram ${LLAMA_CACHE_RAM:-24576}
       --spec-type draft-dflash
-      -md /models/${LLAMA_DRAFT_MODEL_FILE:-Qwen3.8-27B-DFlash2-Q4_K_M.gguf}
+      -md /models/${LLAMA_DRAFT_MODEL_FILE:-Qwen3.8-27B-DFlash2-Q8_0.gguf}
       --spec-draft-n-max ${LLAMA_SPEC_N_MAX:-7}
       --jinja
     # see docs/architecture.md#llama-server (flags), docs/network-access.md#backends (no host port)