improve documentation and example for using pre existing training

[sonalgoyal]

the documentation needs to clearly spell out possible values and structure of the training data, along with adding header to the sample training file.

[tomdavidson]

Thank you! Most of the questions in https://github.com/zinggAI/zingg/discussions/206 are relevant and this could resolve concerns in #200

[sonalgoyal]

b85df565