Python: OpenAI Chat Client consider mapping `max_tokens` to `max_completion_tokens` instead of `max_tokens`

[eavanvalkenburg]

Currently we map max_tokens from ChatOptions to max_tokens for the Chat Completion API, newer models, including reasoning models, use max_completion_tokens instead of max_tokens, we could map to the new field instead, rather then having people do that through additional_properties.

In dotnet, the field in the sdk is already deprecated (called _deprecatedMaxTokens), this is not the case in python, where the field is still shown, it is unclear from docs whether older api versions would break with this change.