Add LLM_HEADERS support for custom request headers

Reads LLM_HEADERS as a JSON object from .env and merges it into every
LLM request alongside the existing Authorization header. Useful for
endpoints that require non-standard headers (e.g. x-openclaw-agent-id).
LLM_API_KEY continues to be sent without the "Bearer" prefix in .env.

Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>

.env.example

@@ -4,8 +4,12 @@ LLM_ENDPOINT=http://localhost:11434/v1/chat/completions
 # Model name passed to the endpoint
 LLM_MODEL=llama3
 
-# Optional API key (sent as Bearer token) — leave blank for local servers
+# Optional API key — value is sent as "Bearer <value>", do NOT include the word "Bearer" here
 LLM_API_KEY=
 
+# Optional extra headers sent with every LLM request, as a JSON object
+# Example: LLM_HEADERS={"x-openclaw-agent-id":"jibo"}
+LLM_HEADERS=
+
 # Default system prompt for the voice AI loop
 LLM_SYSTEM_PROMPT=You are Jibo, a friendly social robot. Keep responses brief and conversational.