generators: update openai for o1 models

[leondz]

The OpenAI "o1" reasoning models have different API expectations from other Chat API models, in terms of params and defaults, and so a new class is added to account for these, and some switching hints placed in the code.

Verification

• List the steps needed to make sure this thing works

export OPENAI_API_KEY=
python -m pytest tests/generators/test_openai.py
python -m garak -m openai -n o1-mini -p test.Test -g1
python -m garak -m openai.OpenAIReasoningGenerator -n o1-mini -p test.Test -g1
python -m garak -m openai.OpenAIReasoningGenerator -n o1-mini -p test.Test --parallel_requests 10

[jmartin-tech]

Suggested refactor is untested, let me know if the pattern does not seem reasonable.

[jmartin-tech]

To support o1 model this also needs to update requirements.txt and pyproject.toml to openai>=1.45.0,<2 as the new max_completion_tokens parameter was added in 1.45.0

[leondz]

thanks, missed that one