Thrift binary fields are not serialized correctly

Thrift binary fields are not serialized correctly - I believe UTF-8 encoding is applied.

A demonstration of this bug is shown in https://github.com/laurencer/parquet-mr-bug

This appears to be an issue in Parquet-MR where Thrift TType is used determine the type of fields. TType actually represents the on-disk/encoded field type tag - that does not distinguish between binary and string fields.

String data is not actually represented on-disk as being different - instead it is up to the program to interpret the binary data as a UTF-8 encoded string. Parquet-MR instead assumes that every binary field is a UTF-8 encoded string.

This may have arisen because the binary field tag is actually TType.String (where it actually just represents a raw binary field).

**Reporter**: [Laurence Rouesnel](https://issues.apache.org/jira/secure/ViewProfile.jspa?name=laurencer)

_{**Note**: *This issue was originally created as [PARQUET-158](https://issues.apache.org/jira/browse/PARQUET-158). Please see the [migration documentation](https://issues.apache.org/jira/browse/PARQUET-2502) for further details.*}

Commencez par la démonstration parquet-mr-bug liée et suivez la gestion par Parquet-MR de Thrift TType.String pour les champs binaires. Confirmez à quel endroit les données binaires sont interprétées comme UTF-8 et comparez ce chemin à celui des champs de type string. Le travail est terminé lorsque les champs binaires bruts sont sérialisés et désérialisés sans interprétation UTF-8, tandis que les champs de type string conservent leur comportement attendu.

Rédigé par le modèle d'indexation à partir du texte de l'issue.