Some fixes on model server

[nehcgs]

1. Improve logging
2. Clean up code and fix tests
3. A little bit refactor
4. Add simple formatting support

[adilhafeez]

Shuguang - thanks for putting this PR together. I've some comments. Please let me know if you have any questions about them.

[adilhafeez]

Thanks for the update, left some comments around logging format. Please take a look at it and let me know if you have any questions.

[adilhafeez]

json.dumps would dump tons of data to disk. Consider trimming it down for example you could do sth line this,

logger.info(f" - [request]: {json.dumps(messages)[:MAX_JSON_LEN]}")

Or better,

def trim_string(input_string, max_length=128):
    if len(input_string) > max_length:
        return input_string[:max_length] + "..."
    return input_string

logger.info(f" - [request]: {trim_string(json.dumps(messages))}")

[nehcgs]

All the data is necessary. Trimming it will have a negative impact on debugging. This is the input to the endpoint, and we should keep it.