Update `collection.batch.rate_limit` to take named vectors into account

[databyjp]

Currently, the rate_limit number does not take named vectors into account. It would be good to have some way of incorporating these.

Because named vectors allow multiple vectorizers, the API might not be super simple.

We might have to provide some way of providing different rate limits for each vectorizer, through init.Integration for example.

[sebawita]

I would prefer to keep rate_limit, as the number of objects added per minute, as calculating rate_limit based on multiple named vectors (with different or same vectorizers) would probably overcomplicate this (now) simple helpful function.

Maybe we could rename the requests_per_minute parameter to objects_per_minute. This way we would make it clearer what is expected to happen.

Mixing vectorizers

Managing batch rate limits across different vectorizers is not a trivial task, and perhaps it could be handled by async vectorization (a different feature we are working on)