[v5] max completion/output token calculations

[gr2m]

Description

Follow up to #5769. @lgrammel talked shortly about this:

packages/openai/src/openai-chat-language-model.ts:

  190        // openai specific settings:
  191:       // TODO remove in next major version; we auto-map maxOutputTokens now
  192        max_completion_tokens: openaiOptions.maxCompletionTokens,

There is a difference between output and completion tokens and it will take some testing and some math to consolidate the options. Or we might end up keeping both. This requires more investigation.