test(e2e): add the gateway proxy config the suite asserts against

tests/e2e/CLAUDE.md documents gateway/ as holding litellm-config.yml, and
litellm-auto's README tells operators to read it, but the directory never
existed. The config the suite actually needs lived only in the deployment
overlay, versioned separately from the tests that depend on it.

That split is why an e2e run cannot currently be tied to a commit: the suite
pins to a SHA while the proxy it exercises floats, and across the eight
retained runs the reported SHA matched the deployed gateway once, drifting
51-73 commits on the rest. Keeping the model list beside the tests lets a run
build both halves from one commit.

Environment-agnostic by construction: every credential is an os.environ/
indirection, and the cache points at REDIS_HOST/REDIS_PORT rather than a
specific endpoint, so the same file works against a bundled Redis or an
external one. Verified by rendering the chart with a bundled database and
Redis - all eight models resolve and coordination_redis is injected.

tests/e2e/gateway/litellm-config.yml

@@ -0,0 +1,74 @@
+general_settings:
+  store_prompts_in_spend_logs: true
+  forward_client_headers_to_llm_api: false
+  maximum_spend_logs_retention_period: 3h
+  maximum_spend_logs_cleanup_cron: 0 1 * * *
+  proxy_config_reload_interval_seconds: 10
+  proxy_budget_rescheduler_min_time: 15
+  proxy_budget_rescheduler_max_time: 20
+litellm_settings:
+  drop_params: true
+  default_redis_ttl: 20
+  json_logs: true
+  store_audit_logs: true
+  cache: true
+  cache_params:
+    type: redis
+    host: os.environ/REDIS_HOST
+    port: os.environ/REDIS_PORT
+  callbacks:
+  - datadog
+  - prometheus
+  - otel
+  require_auth_for_metrics_endpoint: false
+router_settings:
+  fallbacks:
+  - gpt-5.5:
+    - gpt-5.6-luna
+    - gpt-5.2
+  - claude-haiku-4-5:
+    - gemini-2.5-flash
+    - gpt-5.6-luna
+  routing_strategy: simple-shuffle
+  num_retries: 3
+model_list:
+- model_name: gpt-5.5
+  litellm_params:
+    model: openai/gpt-5.5
+    api_key: os.environ/OPENAI_API_KEY
+- model_name: gpt-5.6-luna
+  litellm_params:
+    model: openai/gpt-5.6-luna
+    api_key: os.environ/OPENAI_API_KEY
+- model_name: gpt-5.2
+  litellm_params:
+    model: openai/gpt-5.2
+    api_key: os.environ/OPENAI_API_KEY
+- model_name: claude-haiku-4-5
+  litellm_params:
+    model: anthropic/claude-haiku-4-5
+    api_key: os.environ/ANTHROPIC_API_KEY
+- model_name: azure-foundry-claude-haiku
+  litellm_params:
+    model: azure_ai/claude-haiku-4-5
+    api_base: os.environ/AZURE_AI_API_BASE
+    api_key: os.environ/AZURE_AI_API_KEY
+- model_name: gemini-2.5-flash-vertex
+  litellm_params:
+    model: vertex_ai/gemini-2.5-flash
+    vertex_project: os.environ/VERTEXAI_PROJECT
+    vertex_location: us-central1
+    vertex_credentials: os.environ/VERTEXAI_CREDENTIALS
+- model_name: gemini-2.5-flash
+  litellm_params:
+    model: gemini/gemini-2.5-flash
+    api_key: os.environ/GEMINI_API_KEY
+- model_name: gemini-2.5-flash
+  litellm_params:
+    model: gemini/gemini-2.5-flash
+    api_key: os.environ/GEMINI_API_KEY
+mcp_servers:
+  devin:
+    url: https://mcp.devin.ai/mcp
+    auth_type: api_key
+    auth_value: os.environ/DEVIN_API_KEY