chore(tests): e2e tests should use a fixture for OpenAI client

[alekst23]

Describe what should be investigated or refactored

Currently e2e will cause pytest test discovery to fail. This is because some of the modules cause an API request during the dependency discovery.

Links to any relevant code

client = OpenAI(
    base_url="https://leapfrogai-api.uds.dev/openai/v1", api_key=create_test_user()
)

The create_test_user() function ends up being called during test discovery, not test runs, as it loads dependencies and variables.
This function performs a web request which can fail for a number of reasons, and causes test discovery to fail as well.

Additional context

We can use fixtures to load the client at test run time