Use thread local decoder cache in Binary toStringUsingUTF8()

[asfimport]

In Binary toStringUsingUTF8(), new String(...) is slower because it instantiates a new Decoder, while Charset#decode uses a thread-local decoder cache, it is much faster.

ByteArraySliceBackedBinary is using Charset#decode, while, ByteArrayBackedBinary and ByteBufferBackedBinary are still using new String()

Reporter: Zhenxiao Luo / @zhenxiao
Assignee: Zhenxiao Luo / @zhenxiao

Related issues:

• String decode using 'new String' is slow (duplicates)

_{Note: This issue was originally created as PARQUET-74. Please see the migration documentation for further details.}

[asfimport]

Julien Le Dem / @julienledem:
Thanks for submitting a patch.
This has been merged in
https://github.com/apache/incubator-parquet-mr/pull/40