gemma-4-31B-it: drop context-length and max-running-requests caps

Lets sglang auto-detect gemma4's native context (was capped at 32768,
the source of upstream 400s for long prompts) and removes the explicit
--max-running-requests 64 limit so the scheduler sizes concurrency from
available KV instead.

Author: Evrard-Nil

small-models.yaml

@@ -515,8 +515,6 @@ services:
         --reasoning-parser gemma4
         --tool-call-parser gemma4
         --mem-fraction-static 0.85
-        --max-running-requests 64
-        --context-length 32768
         --chunked-prefill-size 8192
         --num-continuous-decode-steps 5
         --enable-mixed-chunk