Support `dialect.last_comment_rows`

[roll]

Overview

It's been requested multiple times as it's a common practice to have comments at the bottom of csv files.

The implementation should create a buffer in dialect.read_enumerated_content_stream (also see dialect.create_comment_filter) (v5) which has size equal to the comment rows amount.