feat: better cortex serve CLI experience

[Van-QA]

Description:
Running cortex serve returns a 404 error but the error message is not helpful.

Reproduction Steps:

1. Open a terminal.
2. Run cortex serve.
3. Observe the error message

Expected Behavior:
The error message should provide more information on why the command failed.

Environment:
Github codespace
4 cores 16gb RAM

https://discord.com/channels/1107178041848909847/1204617960615055401/1253826168520445996

[0xSage]

I think the problem is that users expect models to be started along with server.

Here's a rough spec for expected behavior:
https://www.notion.so/jan-ai/cortex-serve-2f5293507b094d2abcdf1c0aa809d370?pvs=4

[louis-jan]

Updated: There is Playground URL printed out, also for the error information. Regarding model auto start, it will be confused since we support loading multiple models. It's better to support model param instead. Adding...

[louis-jan]

Regarding whether cortex serve should auto-start models, this is not relevant since it's more about smart routing. We allow users to pull and load multiple models at the same time. It's better to do this with certain requests instead of global serve

e.g.
#803