[server] Update server routes to be compliant with MLServer

[dsikka]

Summary:

• Includes two other PRs: [server] Update OpenAI Model Support #1300 and [server] Refactor + OpenAI Chat Completion Support #1288
• Updates the base routes such that they follow the mlserver convention

Testing

• Updated all the server tests
• Tested locally using custom routes as well as routes built for the user given the model name

Sample Config:

num_cores: 2
num_workers: 2
endpoints:
  - task: question_answering
    model: zoo:nlp/question_answering/bert-base/pytorch/huggingface/squad/12layer_pruned80_quant-none-vnni
    route: some_route/pruned

This will now produce the following endpoints:

[bfineran]

LGTM overall - as discussed offline, agree with both @dsikka and @Satrat it's time we move route creation to a separate file / fleshed out system

[dbogunowicz]

LGTM, clean

[bfineran]

LGTM - as mentioned before we need to sync w/ QA before landing

[mgoin]

This seems like it would cause breaking changes to any application built using the old endpoint structure. I think if you could make a README or document showing how to transition current examples of usage, that would be helpful for other teams and users to have a summary of the changes.
For instance, how should we update this Digital Ocean getting started guide? https://marketplace.digitalocean.com/apps/deepsparse-inference-runtime