This breaks some of the consumer operations into request generation /
response parsing methods.

The public API does not change. However, this allows power users who are
willing to deal with risk of private methods changing under their feet
to decouple generating the message futures from processing their
responses. In other words, you can use these to fire a bunch of request
at once and delay processing the responses until all requests are fired.

Delaying the conversion to seconds makes the code intent more clear.

The docs for `api_version_auto_timeout_ms` mention setting
`api_version='auto'` but that value has been deprecated for years in
favor of `api_version=None`.

Updating the docs for now, and will remove support for `'auto'` in next
major version bump.

If the cluster metadata object has no info about the topic, then issue a blocking metadata call to fetch it.