CWM data contains double-encoded unicode

[wu-lee]

Problem

Some of the data in the delhi demo has double-enoded text.

For instance, see "Cooperativa de Ahorro y Créditos Norte Grande"

Resolution

We would need to sweep the data for these mojibake issues. The corrections will need to be traced to source (probably upstream) and corrected there.

We might also add an automated check for known mojibake cases, these will be identifiable by characteristic junk characters as above.

Acceptance criteria

• An automated check exists.
• An (automated?) sweep of the data returns no cases.