Provide an API to retrieve the number of requests being processed

[cr7258]

We have integrated Ollama into our inference platform, and we are currently implementing a feature that waits for all active requests to complete before shutting down the pod, ensuring a graceful termination.

We hope Ollama can provide an API for retrieving the number of requests being processed, this could be a Prometheus metrics (gauge type), for example:

• llama.cpp metrics
• vLLM metrics

[googs1025]

/cc

[rick-github]

#3144

[ParthSareen]

Will close this in favor of tracking https://github.com/ollama/ollama/issues/10419

[hendrikebbers]

@ParthSareen I assume you mean #3144 ?