Add unbuffer support to GoogleHadoopFSInputStream

[stiga-huang]

It's an improvement for GoogleHadoopFSInputStream to implement the CanUnbuffer interface. The unbuffer() operation instructs the source to release any system resources they are currently holding on to, such as buffers, sockets, file descriptors, etc. Any subsequent IO operation will likely have to reacquire these resources. Unbuffering is useful in situation where streams need to remain open, but no IO operation is expected from the stream in the immediate future (examples include file handle cacheing).

Hadoop doc: https://hadoop.apache.org/docs/r3.3.0/hadoop-project-dist/hadoop-common/filesystem/fsdatainputstream.html#CanUnbuffer.unbuffer.28.29

Related issues for other storages:

• S3: https://issues.apache.org/jira/browse/HADOOP-14747
• ABFS: https://issues.apache.org/jira/browse/HADOOP-16859