Parquet-tools cat mangles strings created by other clients

I have some parquet files that are created by Java MR process (which I do not own). I am able to read these fields successfully in pig and Spark, but for some reason the String fields are being mangled when I view the files with parquet-tools (cat).

Here are the details on the file metadata using today's build of parquet-tools:
```

hadoop jar parquet-tools-1.9.1-SNAPSHOT.jar meta /parquet-r-00000
```
Output:
```

file: hdfs:///parquet-r-00000
creator: parquet-mr version 1.8.1 (build 4aba4dae7bb0d4edbcf7923ae1339f28fd3f7fcf)

file schema: MY_DATA
--------------------------------------------------------------------------------
myfield: OPTIONAL BINARY R:0 D:1

row group 1: RC:37343 TS:32397576 OFFSET:4
--------------------------------------------------------------------------------
myfield: BINARY SNAPPY DO:0 FPO:4 SZ:273374/556406/2.04 VC:37343 ENC:RLE,BIT_PACKED,PLAIN_DICTIONARY ST:[no stats for this column]
```
 Has anyone seen this before?

**Environment**: {noformat}
uname -a
Linux myhost 4.4.0-63-generic #84-Ubuntu SMP Wed Feb 1 17:20:32 UTC 2017 x86_64 x86_64 x86_64 GNU/Linux
{noformat}
**Reporter**: [Michael McCarthy](https://issues.apache.org/jira/secure/ViewProfile.jspa?name=Michael.McCarthy)

_{**Note**: *This issue was originally created as [PARQUET-1235](https://issues.apache.org/jira/browse/PARQUET-1235). Please see the [migration documentation](https://issues.apache.org/jira/browse/PARQUET-2502) for further details.*}

Comienza con la ruta de parquet-tools cat y los metadatos del campo OPTIONAL BINARY descrito en el informe, y luego compara su salida con las cadenas leídas por Pig y Spark. Reproduce la alteración usando un archivo Parquet creado por el proceso Java MR; el trabajo estará terminado cuando parquet-tools cat muestre correctamente esos campos de cadena.

Escrito por el modelo de indexación a partir del texto del issue.