feat: DIA-1362: Add custom LLM endpoint testing scripts

[niklub]

This PR adds tests to verify Custom LLM endpoint connection.

How to create a custom LLM endpoint

1. Download and create ollama server
2. Run local server with ollama run llama3.1
3. Connect your instance using the following parameters (note /v1 suffix):
   • endpoint: http://localhost:11434/v1
   • deployment name: llama3.1
   • api_key: ollama
4. (optional) To connect remotely (i.e. when localhost is not available to use), use ngrok or any other proxy server
5. (optional) To enforce authentication header, use one of the following option:
   • specifying Basic Auth with ngrok. Not that after specifying user:pass string, it must be provided as base64 encoded parameter in the corresponding custom LLM endpoint field. For example, run echo -n "user:pass" | base64, and then use the string: Basic <base64 encoded credentials>
   • use auth_proxy_server.py script launching it with: TARGET_URL=http://localhost:11434 EXPECTED_HEADER=secret uvicorn auth_proxy_server:app and provide secret as Authentication header field in Custom LLM endpoint connection

[ellipsis-dev]

Add a check to ensure TARGET_URL is set. If not, raise an exception or log an error to prevent runtime issues.

[ellipsis-dev]

Add assertions to verify the expected behavior of the agent. Currently, the test only prints results without checking for correctness.

[robot-ci-heartex]

Commit https://github.com/HumanSignal/infra/commit/a04fbaf875663e567c580b77b90874f1ecefc9b6 is created.

[robot-ci-heartex]

Commit https://github.com/HumanSignal/infra/commit/af988e965e349e75fafdfd0b375068923aba7cd3 is created.