Fix parallel validation, remove ngram-mod

- Rewrite validation script with temp files for true parallel curl requests
- Both prompts now complete 1024 tokens each (was cutting first short)
- Remove ngram-mod from spec pipeline (was -1.2 tok/s regression)
- True parallel throughput: 15.68 tok/s across 2 streams

scripts/serve-llamacpp-dflash.sh

@@ -26,7 +26,7 @@ echo ""
 exec "${SERVER_BIN}" \
     --hf-repo "${TARGET_REPO}" \
     --spec-draft-hf "${DRAFT_REPO}" \
-    --spec-type draft-dflash,ngram-mod \
+    --spec-type draft-dflash \
     --spec-draft-n-max "${DRAFT_N_MAX}" \
     --device ROCm0 \
     --spec-draft-device ROCm0 \