Standardize text content handling into new textContent AS1 field for index-based spans in tags

[snarfed]

Right now, our code for handling HTML vs text content is a bit of a mess. For example, extracting and converting links, mentions, hashtags, and other entities in HTML content into tags is in Source.postprocess_object (mentions), bluesky.from_as1 (links and hashtags?), and some other places too.

We should centralize all this into a new AS1 field for text content. Ideally something existing in AS1, but if not, maybe a new textContent field. content would always be HTML, everything would be in markup there. tags with startIndex/length would always reference textContent, we could reliably convert plain text whitespace (eg newlines) to <br>s, and we could get rid of content_is_html.

Background: #675, snarfed/bridgy-fed#990, etc