Generalize LLM backend configuration

[eshulman2]

Summary

• replace product-facing Claude-specific wording with Deep Agents/LangChain model backend language
• add generic LLM_API_KEY, VERTEX_PROJECT_ID, and VERTEX_REGION config aliases while preserving ANTHROPIC_* compatibility
• keep the default orchestrator/container model direct-Anthropic compatible with claude-sonnet-4-5@20250929, while documenting Gemini as a Vertex AI option
• add tests covering generic/legacy LLM config resolution and the default model

Validation

• UV_CACHE_DIR=/tmp/uv-cache uv run ruff check src/forge/config.py src/forge/integrations/agents/agent.py containers/entrypoint.py src/forge/sandbox/runner.py tests/unit/test_config_prd.py
• UV_CACHE_DIR=/tmp/uv-cache uv run pytest tests/unit/test_config_prd.py
• runtime check: with LLM_API_KEY set and LLM_MODEL unset, ForgeAgent creates ChatAnthropic