Consider refactoring string escape errors

[dmarcotte]

We currently have some code in Lexer specifically designed to serve validating string escapes and reporting precisely located errors on bad escapes (for instance, here). This works well and has not caused problems, so likely it does not need attention, but it worth calling out that the new sublocation utilities implemented in #286 may allow for a cleaner approach to this now. If/when the current approach causes problems, we should consider this refactor.

[dmarcotte]

Refactored this in #305

Some notes on the ideas in this original issue:

• I mentioned this "likely does not need attention", and that's almost still true. Doing the refactor now was motivated by wanting to reuse these validations as part of better formalizing embed tags as proper KSON strings, ensuring proper JSON compatibilty. That work is forthcoming
• the sublocation work actually cannot be used here since this validation happens lower-level in the parse, inspecting the raw KSON strings to see if they are valid enough to be used in the subparser sublocation system