ProtoReader does not iterate over the parquet file correctly.

The `ProtoParquetReader` does not iterate over the parquet file correctly, but it gets stuck in the first element and keeps reading as many times as elements the file contained.

In my Scala example I am just reading from a local file that I know for sure it contains right data.

```

val hadoopCOnf = new Configuration()

val outfile: String = genTemporaryFile()

val r: ParquetReader[Event.Builder] = {
ProtoParquetReader.builder[Event.Builder](new Path(outfile)).withConf(hadoopCOnf).build()
}

```

Notice that the proto schema that I am using is generated from 

The generated proto implements com.google.protobuf.GeneratedMessageV3.

See an example on how the ProtoParquetReader is created line(65 and 69): 

and here how it is used (notice that is defined for only one record, the same one for multiple records would fail) 

**Reporter**: [Pau Alarcon ](https://issues.apache.org/jira/secure/ViewProfile.jspa?name=paualarco)

_{**Note**: *This issue was originally created as [PARQUET-1871](https://issues.apache.org/jira/browse/PARQUET-1871). Please see the [migration documentation](https://issues.apache.org/jira/browse/PARQUET-2502) for further details.*}

ProtoParquetReader から始め、monix-connect の ProtoParquetFixture.scala の 65～69 行付近での構築方法と、ProtoParquetSpec.scala の 83 行付近での使用方法を比較します。複数のレコードで動作を再現し、反復処理によって最初のレコードが繰り返されるのではなく、後続のレコードが返されることを確認します。

索引モデルが issue の本文から書いたものです。