Support configurable for DirectByteBufferAllocator from Hadoop Configuration

Now in [HadoopReadOptions|([https://github.com/apache/parquet-mr/blob/master/parquet-hadoop/src/main/java/org/apache/parquet/HadoopReadOptions.java#L85]] class, we cannot change default allocator from Hadoop Configuration.

Add a config parquet.allocator.direct.enabled to enable DirectByteBufferAllocator in ParuqetFileReader. 

Reporter: ShuMing Li

Related issues:

• ColumnChunkPageWriter uses only heap memory. (relates to)

PRs and other links:

• GitHub Pull Request #749

_{Note: This issue was originally created as PARQUET-1771. Please see the migration documentation for further details.}

从 parquet-hadoop/src/main/java/org/apache/parquet/HadoopReadOptions.java 以及 ParquetFileReader 中选择 allocator 的路径开始。跟踪 Hadoop Configuration 的读取方式，以及 DirectByteBufferAllocator 的选择方式；当 parquet.allocator.direct.enabled 控制该选择，并且未启用时保留现有行为，即表示完成。