Localize (i.e., translate) strings emitted by built-in converters and document how this feature works.

We'll need two things. First, we need to make sure that any and all strings that are emitted read from an attribute. This is true in most cases, such as note-caption. Next, we need to decide the best way to locate and read in attributes based on language selection. While these strings can be fed in one by one into the attributes, it's rather inconvenient and does not encourage reuse. We want to eventually have support for many languages inside of Asciidoctor like AsciiDoc Python.