Add recommended sampling defaults and reduce MoE GPU offload on dankserver

- frank-default-large: add TEMP=1.0, TOP_K=20, TOP_P=0.95, MIN_P=0.0
  (per HauhauCS Qwen3.6-27B model card, thinking-general mode)
- frank-default-small: add same sampling defaults
  (per HauhauCS Qwen3.6-35B-A3B model card)
- frank-default-small: reduce N_CPU_MOE from 33 to 34 (6 expert
  layers on GPU instead of 7) to prevent OOM during high-load sessions
- Both serve scripts pass --temp, --top-k, --top-p, --min-p to
  llama-server; clients can still override per-request via API

scripts/serve-llama-frank-small.sh

@@ -32,12 +32,16 @@ fi
 : "${NGRAM_MOD_N_MATCH:=24}"
 : "${CACHE_TYPE_K:=q8_0}"
 : "${CACHE_TYPE_V:=f16}"
-: "${N_CPU_MOE:=40}"     # all 40 layers: MoE expert weights stay on CPU
+: "${N_CPU_MOE:=34}"     # 34 of 40 layers: MoE expert weights stay on CPU; 6 on GPU
 : "${N_GPU_LAYERS:=all}" # every non-expert tensor goes to GPU
 : "${BATCH_SIZE:=2048}"
 : "${UBATCH_SIZE:=512}"
 : "${CTX_CHECKPOINTS:=0}"
 : "${NO_MMAP:=0}"
+: "${TEMP:=1.0}"
+: "${TOP_K:=20}"
+: "${TOP_P:=0.95}"
+: "${MIN_P:=0.0}"
 : "${THREADS:=8}"        # 7700X physical cores
 : "${THREADS_HTTP:=8}"
 
@@ -70,6 +74,10 @@ exec "$LLAMA_SERVER" \
     --ubatch-size "$UBATCH_SIZE" \
     --ctx-checkpoints "$CTX_CHECKPOINTS" \
     $([ "$NO_MMAP" = "1" ] && printf -- "--no-mmap") \
+    --temp "$TEMP" \
+    --top-k "$TOP_K" \
+    --top-p "$TOP_P" \
+    --min-p "$MIN_P" \
     --threads "$THREADS" \
     --threads-http "$THREADS_HTTP" \
     --alias "$ALIAS" \