fix: sgml parsing edge case when word list delimiter in transcript

  - SCTK unfortunately doesn't escape the quote the entire field in the
    sgml file, and causes encoding/csv to split at the wrong place for
    a string like `S,"hello:","hello":C,"how","how"` with delimiter = `:`.

  - This commit applies a fix where we change the delimiter to an unlikely
    string using a carefully constructed regex, and them split on the new
    unlikely string.

doc: update README example

doc: add README with simple usage example