Add server stand up test as part of MLX CI

[Tomcli]

Hi @ckadner sorry I didn't have time to recompile the server for every PR. Right now our CI is only catching build errors but we don't have any test to catch runtime errors. Maybe we can add a simple test just to start the server to make sure nothing is crashing the server startup? This way we don't accidentally merged critical bugs that crashes our server startup.

Originally posted by @Tomcli in https://github.com/machine-learning-exchange/mlx/issues/205#issuecomment-920338732

[ckadner]

@Tomcli -- so our PR builder could do this for API code changes:

• [x] build a new API docker image, push it
• [ ] pull the new image image and start it -- OR -- run the Python code directly
• [ ] with Docker Compose: upload the catalog via curl, check response code (#23)
• [ ] with Docker Compose: get the catalog, check response code, check number of assets is greater 0

Besides that, I still need to create unit tests from the examples I usually run as a manual test

• [ ] Related: #116

[Tomcli]

@Tomcli -- so our PR builder could do this for API code changes:

• [x] build a new API docker image, push it
• [ ] pull the new image image and start it -- OR -- run the Python code directly
• [ ] with Docker Compose: upload the catalog via curl, check response code (Rework bootstrapper #23)
• [ ] with Docker Compose: get the catalog, check response code, check number of assets is greater 0

Besides that, I still need to create unit tests from the examples I usually run as a manual test

• [ ] Related: Create API unittests with Docker Compose #116

I was more thinking of a simple test just start the server in a Github Action, so we don't get a crashing image in our latest build. The docker compose test is nice to have, but I'm not sure can github action able to handle that many workloads.

[ckadner]

starting the image after building it should be a good start -- that does not need Minio or MySQL