Expand OPENAI_MODELS list with recent OpenAI model names

[Cyberfilo]

Summary

`OPENAI_MODELS` in `aider/models.py` is used by `sanity_check_models` to map a bare OpenAI model name (no provider prefix) to the `OPENAI_API_KEY` environment variable. Models not in this list fall through to the keymap provider lookup, which fails when the user passes the name without a provider prefix.

The list was stale: it had `o1`, `o1-preview`, `o1-mini`, `o3-mini`, the `gpt-4*` family, `gpt-4o*`, `gpt-3.5-*`, plus `gpt-5.5` and `gpt-5.5-pro` — but was missing:

• The entire `o3` / `o3-pro` / `o4-mini` reasoning series.
• The `gpt-4.1` family (4.1, 4.1-mini, 4.1-nano).
• All `gpt-5` / `gpt-5.1` / `gpt-5.2` / `gpt-5.3` / `gpt-5.4` family variants (chat / pro / mini / nano / codex).

All added names already have corresponding entries in `aider/resources/model-settings.yml`, so the list now matches the rest of the project.

Effect

Before this change, a user running `aider --model gpt-5.4-mini` with `OPENAI_API_KEY` set in the environment would see "OPENAI_API_KEY: Not set" because the bare name didn't match the OPENAI_MODELS string list. After this change, the API key is correctly detected.

Test plan

• `python -c "from aider.models import OPENAI_MODELS; print(len(OPENAI_MODELS), 'o4-mini' in OPENAI_MODELS, 'gpt-5.4-mini' in OPENAI_MODELS, 'gpt-4.1' in OPENAI_MODELS)"` -> `61 True True True`
• Diff is purely additive: 30 lines added to the OPENAI_MODELS string + 1 line in HISTORY.md.