Improve representation of redundant information in RichValue

Currently, messages are parsed into (usually compound) `RichValue`s containing _both_ the raw text and more structured information like mentions, urls, ... This duplication is usually not a problem, since most commands only request a single representation, however e.g. when outputting the message again, for example using `IdentityCommand`, the structured values may be output multiple times.

Maybe a more structured representation would be better for `RichValue`s, e.g. a document-esque structure that holds the unstructured (raw text) representation in a separate attribute?

Start by tracing the RichValue representation used when messages are parsed and then passed to IdentityCommand. Examine how raw text, mentions, and URLs are stored and rendered. Done would mean an agreed representation that avoids structured values being output multiple times, but the issue does not identify files or tests.

Written by the indexing model from the issue text.