clarify that character references are not interpreted in HTML blocks and raw HTML

[rsc]

The wording in section 2.5 implied otherwise, although an example in 6.6 did clarify. Adjust the first to remove the false implication and adjust the second from saying "preserved", which might mean they have to be parsed first, to "copied uninterpreted".

The diff makes a bit of a hash of the changes, which are to move the "unrecognized" cases above the recognized ones, as suggested in #687.

Fixes #687.