fix: remove hardcoded max_tokens/temperature from _dispatch_via_llm

Both values are now sourced from team.yaml (models.default_max_tokens and
models.default_temperature) via the adapter's __init__, eliminating the
last hardcoded magic numbers. Callers can still override per-call via
context dict if needed.

core/team_runner.py

@@ -328,8 +328,6 @@ class TeamRunner:
             capability=capability,
             context={
                 "system_prompt": system_prompt,
-                "max_tokens": 4096,
-                "temperature": 0,
             },
         )
         return self._extract_json(raw)