Parquet pages for different columns cannot be read in parallel

All ColumnChunkPageReader instances use the same decompressor. 

```java

BytesInputDecompressor decompressor = options.getCodecFactory().getDecompressor(descriptor.metadata.getCodec());
return new ColumnChunkPageReader(decompressor, pagesInChunk, dictionaryPage);
```
The CodecFactory caches the decompressors for every codec type returning the same instance on every getCompressor(codecName) call. See the caching happening here:

```java

@Override
public BytesDecompressor getDecompressor(CompressionCodecName codecName) {
BytesDecompressor decomp = decompressors.get(codecName);
if (decomp == null){
decomp = createDecompressor(codecName); decompressors.put(codecName, decomp);
}
return decomp;
}
 
```
 

If multiple threads try to read the pages belonging to different columns, they run into thread

safety issues. This issue prevents increasing the throughput at which applications can read parquet data by parallelizing page reads. 

**Reporter**: [Samarth Jain](https://issues.apache.org/jira/secure/ViewProfile.jspa?name=samarthjain) / @samarthjain
**Assignee**: [Samarth Jain](https://issues.apache.org/jira/secure/ViewProfile.jspa?name=samarthjain) / @samarthjain
#### PRs and other links:
- [GitHub Pull Request #668](https://github.com/apache/parquet-mr/pull/668)
- [GitHub Pull Request #670](https://github.com/apache/parquet-mr/pull/670)

_{**Note**: *This issue was originally created as [PARQUET-1641](https://issues.apache.org/jira/browse/PARQUET-1641). Please see the [migration documentation](https://issues.apache.org/jira/browse/PARQUET-2502) for further details.*}

Commencez par ParquetFileReader.java vers la ligne 1286 et CodecFactory.java vers la ligne 197 afin de suivre la manière dont les décompresseurs sont partagés par les instances de ColumnChunkPageReader. Consultez les pull requests #668 et #670 avant de commencer, puis vérifiez que les lectures de pages pour différentes colonnes peuvent s’exécuter en parallèle sans le problème de sécurité des threads signalé.

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