fix(litellm): route kimi-k3 to the Kimi Code (coding) endpoint

The Heid panel plan uses Kimi's coding endpoint, not the general Moonshot API.
kimi-k3 now → openai/k3 @ https://api.kimi.com/coding/v1 (KIMI_CODE_API_KEY,
Vivace); the original general-endpoint entry is kept as kimi-k3-gen-api
(api.moonshot.ai, MOONSHOT_API_KEY). Both verified live through the gateway.
Same k3 constraints on both: temperature MUST be 1 (else 400), reasoning model
(reasoning_content vs content, needs adequate max_tokens).

stacks/litellm/compose.yaml

@@ -50,6 +50,10 @@ services:
       # Moonshot Kimi, etc.). Paid — only gateway-keyed callers reach them,
       # but they spend.
       - Z_AI_API_KEY=${Z_AI_API_KEY:-}
+      # Kimi: KIMI_CODE_API_KEY = the coding endpoint (api.kimi.com/coding, the
+      # primary kimi-k3 arm); MOONSHOT_API_KEY = the general api.moonshot.ai
+      # endpoint (the kimi-k3-gen-api variant).
+      - KIMI_CODE_API_KEY=${KIMI_CODE_API_KEY:-}
       - MOONSHOT_API_KEY=${MOONSHOT_API_KEY:-}
       # Langfuse-ready: blank until you bolt Langfuse on. Filling these +
       # uncommenting the callback in config.yaml is the entire upgrade.