It turns out it's really useful to have the offset (position in the document if it were a stream of chars) of a bug. Required, in fact, for some applications to figure out where in a document something is.

We can figure this out via line/col by having things keep tables of line/offset positions, but generally they'd need the source input to build this table, and thus another hit on the file system and whatnot (which is rough with lots of files).

The parser already has this info, so we should expose it and save any tools that use offsets versus line+column the trouble of converting the location info. This also adds a length attribute, which is a little hokey but seems to work fine (used to visually highlight the code causing the issue).