STAC API and Tiler not using gunicorn worker and worker configuration

I just noticed that the Tiler API and STAC APIs are launching uvicorn directly to start FastAPI and not using gunicorn with a server worker.

This is fine for development, but for a production FastAPI instance it would create a queueing bottleneck for concurrent requests.

Secondly, there is no default worker processes and/or thread configuration for GUnicorn configured. This would significantly improvement throughput of the application (like a 4x improvement in requests/sec).

See https://www.uvicorn.org/deployment/#using-a-process-manager and https://fastapi.tiangolo.com/deployment/server-workers/

Spotted this in these places:

https://github.com/microsoft/planetary-computer-apis/blob/main/pctiler/Dockerfile#L47

https://github.com/microsoft/planetary-computer-apis/blob/main/pcstac/Dockerfile#L17

https://github.com/microsoft/planetary-computer-apis/blob/main/docker-compose.yml#L19

https://github.com/microsoft/planetary-computer-apis/blob/main/docker-compose.yml#L38

Happy to submit a PR with benchmark, unless I'm missing something

Start by reviewing pctiler/Dockerfile, pcstac/Dockerfile, and the two service entries in docker-compose.yml, then compare their startup commands with the Uvicorn and FastAPI deployment guidance linked in the issue. Done means the production API services use the intended process manager and have a defined worker configuration; benchmarks may be added if they are part of the agreed scope.