updated tests

[shreymodi1]

Note

Updates streaming compliance benchmarks and rollout plumbing.

• Default model set to glm-4p6; import pytest and add @pytest.mark.skipif to conditionally run reasoning and multi-tool tests via EP_SUPPORTS_REASONING and EP_SUPPORTS_MULTIPLE_TOOL_CALLS
• New helpers _maybe_add_reasoning_effort and updated _build_completion_params_from_payload to only pass reasoning_effort when supported; many tests now use these helpers
• Add raw_output: True to test params; rollout moves raw_output to extra_body and captures response raw_output into row.execution_metadata.raw_output
• Model updates: ExecutionMetadata gains raw_output field to store provider extras
• Tests adjusted to include raw_output, and make reasoning checks/metrics conditional on support; minor param tweaks across streaming and non-streaming cases

^{Written by Cursor Bugbot for commit da015c5. This will update automatically on new commits. Configure here.}

[dphuang2]

lets avoid test-specific environment variables in favor of globally configurable environment variables

[dphuang2]

please add the new flags in https://github.com/eval-protocol/python-sdk/blob/main/eval_protocol/pytest/plugin.py