[SPARK-9460] Avoid byte array allocation in StringPrefixComparator.

[rxin]

As of today, StringPrefixComparator converts the long values back to byte arrays in order to compare them. This patch optimizes this to compare the longs directly, rather than turning the longs into byte arrays and comparing them byte by byte (unsigned).

This only works on little-endian architecture right now.

[rxin]

cc @tellison can you take a look at this (or find somebody else) to see if we can fix it by just not doing the reverse bytes in big-endian?

[davies]

We should check the numBytes here, or get random bytes if it's less than 8 bytes.

[davies]

We should also test the endian-ness.

[rxin]

I created https://github.com/apache/spark/pull/7789/files to make it safer. Still doesn't handle big endian though.

[rxin]

cc @davies / @JoshRosen

[rxin]

cc @JoshRosen

I made the randomized tester to be stricter. Previously it would still pass if prefix always return 0.

[SparkQA]

Test build #38943 has finished for PR 7765 at commit 4c8d094.

• This patch fails PySpark unit tests.
• This patch merges cleanly.
• This patch adds no public classes.

[SparkQA]

Test build #38950 has finished for PR 7765 at commit e4908cc.

• This patch passes all tests.
• This patch merges cleanly.
• This patch adds no public classes.