Add a possibility to distinguish between chat and complete LM requests

[dhuebner]

Feature Description:

Sometimes it is important to know if a LanguageModelRequest is called from the chat or from the editor. For example for completion requests one would normally don't use streams as response, also sometimes a post-process step is requiered in some cases.

Possible solutions:

• additional property requestKind or kind
• new interface (may benefit from additional completion related properties; don't need messages property)

It would probably be even more beneficial to have an additional method completeionRequest(...) in the LanguageModel interface as the request and response are often differs a lot.