perf(frank-small): cap the CPU RAM cache at 2048 MiB

The build default is 8192 MiB. dankclaw is memory-tight (frank-large's
70 GiB GTT plus the rest of the stack) and Gemma4 runs fully on the 3090
(-ngl 999), so the CPU-side cache is not the hot path. Cap it at 2048 to
give back ~6 GiB of LPDDR. Gemma4 is not a DeltaNet model, so the
>12.8 GiB CACHE_RAM floor does not apply.

Wired through frank-small.env (CACHE_RAM) -> serve-frank-small.sh
(--cache-ram), with a wiring seam so the cap cannot be silently unwired.

scripts/serve-frank-small.sh

@@ -58,6 +58,7 @@ fi
 : "${PARALLEL:=1}"
 : "${CACHE_TYPE_K:=f16}"
 : "${CACHE_TYPE_V:=f16}"
+: "${CACHE_RAM:=8192}"
 : "${BATCH_SIZE:=2048}"
 : "${UBATCH_SIZE:=512}"
 # Decode-shape knobs: the MTP drafter's per-round proposal length (build
@@ -130,6 +131,7 @@ exec "$LLAMA_BIN" \
     --poll "$POLL_LEVEL" \
     --cache-type-k "$CACHE_TYPE_K" \
     --cache-type-v "$CACHE_TYPE_V" \
+    --cache-ram "$CACHE_RAM" \
     "${SPEC_ARGS[@]}" \
     --jinja \
     --metrics \