ParquetMetadata.fromJSON method doesn't work

I wrote a simple testCase to test the fromJSON method below and it fails because ParquetMetadata does not implement default constructors.

```

@Test public void testFromJSON() {
MessageType schema = parseMessageType("message test { optional binary some_null_field; }");
parquet.hadoop.metadata.FileMetaData fileMetaData =
new parquet.hadoop.metadata.FileMetaData(schema, new HashMap(), null);
List blockMetaDataList = new ArrayList();
BlockMetaData blockMetaData = new BlockMetaData();
blockMetaData.addColumn(createColumnChunkMetaData());
blockMetaDataList.add(blockMetaData);
ParquetMetadata metadata = new ParquetMetadata(fileMetaData, blockMetaDataList);

String metadataStr = ParquetMetadata.toJSON(metadata);
System.out.println(ParquetMetadata.toPrettyJSON(metadata));
ParquetMetadata m2 = ParquetMetadata.fromJSON(metadataStr);
}
```

**Reporter**: [Vihang Karajgaonkar](https://issues.apache.org/jira/secure/ViewProfile.jspa?name=vihangk1)

_{**Note**: *This issue was originally created as [PARQUET-851](https://issues.apache.org/jira/browse/PARQUET-851). Please see the [migration documentation](https://issues.apache.org/jira/browse/PARQUET-2502) for further details.*}

Start at ParquetMetadata.fromJSON and the related ParquetMetadata.toJSON call shown in the reproduction, then run the supplied test case with FileMetaData and BlockMetaData. Done means metadata serialized by toJSON can be passed to fromJSON without failing because of missing default constructors.

Written by the indexing model from the issue text.