Clean up implementations of LLM Service

[mdemoret-nv]

The LLM Service wrappers have evolved over time and need to be cleaned up to be uniform. The following items should be examined and updated to be uniform across the 3 implementations: "openai", "nvfoundation", "nemo". If any other types have been added, they should be cleaned up as well.

• Services should handle environment variables using the new EnvConfigValue class
• All arguments for common methods should use similar names
• All docstrings should be similar and accurate
• Tests for these wrappers should be similar and laid out consistently
  • Only one test file per service (remove any service/client separate test files)
  • Tests should exercise creation of the class via the LLMService.create function
• Test methods should be arranged in a similar order
• Each service should be used in the tests/llm/test_*_pipe.py tests