Return finish reason field & VLLM Refactor

[CollectiveUnicorn]

When performing a completion or chat completion return the finish_reason according to the api spec.

• Adds finish reason in completion/chat completion that mirrors the openai spec
• Adds token usage in completion/chat completion responses that mirrors the openai spec

Closes https://github.com/defenseunicorns/leapfrogai/issues/463 Closes https://github.com/defenseunicorns/leapfrogai/issues/431

[netlify[bot]]

✅ Deploy Preview for leapfrogai-docs canceled.

Name	Link
🔨 Latest commit	99661575447dd4871a298155cf60c73dde288cc0
🔍 Latest deploy log	https://app.netlify.com/sites/leapfrogai-docs/deploys/664cd436bf4db50008143d93

[CollectiveUnicorn]

Conforming to the specs for token length will require updating all of the backends. It assumes that there's the token counting function. TBD if doing that in this commit.