remove max_tokens from request in case of o3 and o1 models

[narengogi]

o3 and o1 models do not support max_tokens, they expect max_completion_tokens instead which includes all the tokens used for reasoning also. though I dont think it is right to map max_tokens to max_completion_tokens incase the user only sends max_tokens and max_completion_tokens, we should atleast remove max_tokens so that it doesn't throw an error.

Since max_tokens is depecated in favour of max_completion_tokens we should just map it to max_completion_tokens