fix: propagate alternative API provider settings to agent subprocesses

When users configured GLM/Ollama/Kimi via the Settings UI, agents still
used Claude because conflicting env vars leaked through subprocess env.

Root cause: get_effective_sdk_env() set ANTHROPIC_AUTH_TOKEN for GLM but
didn't clear ANTHROPIC_API_KEY, which leaked from os.environ. The CLI
prioritized the wrong credential.

Changes:
- registry.py: Clear conflicting auth vars (API_KEY vs AUTH_TOKEN) and
  Vertex AI vars when building env for alternative providers
- client.py: Replace manual os.getenv() loop with get_effective_sdk_env()
  so agent SDK reads provider settings from the database
- autonomous_agent_demo.py: Apply UI-configured provider settings to
  process env so CLI-launched agents also respect Settings UI config
- start.py: Pass --model from settings when launching agent subprocess
- server/schemas.py: Allow non-Claude model names when an alternative
  provider is configured (prevents 422 errors for glm-4.7, etc.)
- .env.example: Document env vars for GLM, Ollama, and Kimi providers

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

start.py

@@ -390,8 +390,11 @@ def run_agent(project_name: str, project_dir: Path) -> None:
     print(f"Location: {project_dir}")
     print("-" * 50)
 
-    # Build the command - pass absolute path
-    cmd = [sys.executable, "autonomous_agent_demo.py", "--project-dir", str(project_dir.resolve())]
+    # Build the command - pass absolute path and model from settings
+    from registry import DEFAULT_MODEL, get_all_settings
+    settings = get_all_settings()
+    model = settings.get("api_model") or settings.get("model", DEFAULT_MODEL)
+    cmd = [sys.executable, "autonomous_agent_demo.py", "--project-dir", str(project_dir.resolve()), "--model", model]
 
     # Run the agent with stderr capture to detect auth errors
     # stdout goes directly to terminal for real-time output