[SPARK-9206] [SQL] Fix HiveContext classloading for GCS connector.

[dennishuo]

IsolatedClientLoader.isSharedClass includes all of com.google., presumably
for Guava, protobuf, and/or other shared Google libraries, but needs to
count com.google.cloud. as "hive classes" when determining which ClassLoader
to use. Otherwise, things like HiveContext.parquetFile will throw a
ClassCastException when fs.defaultFS is set to a Google Cloud Storage (gs://)
path. On StackOverflow: http://stackoverflow.com/questions/31478955

EDIT: Adding @yhuai who worked on the relevant classloading isolation pieces.

[marmbrus]

ok to test

[SparkQA]

Test build #37878 has finished for PR 7549 at commit 1f8db07.

• This patch fails Spark unit tests.
• This patch merges cleanly.
• This patch adds the following public classes (experimental):
  • trait Nondeterministic
  • abstract class RDG extends LeafExpression with Nondeterministic
  • case class Rand(seed: Long) extends RDG
  • case class Randn(seed: Long) extends RDG

[marmbrus]

test this please

[marmbrus]

test this please

[SparkQA]

Test build #37970 has finished for PR 7549 at commit 1f8db07.

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

[marmbrus]

Thanks! Merging to master.