[BACKPORT-2.2][SPARK-19372][SQL] Fix throwing a Java exception at df.fliter() due to 64KB bytecode size limit

[kiszk]

What changes were proposed in this pull request?

This PR is backport of #17087 to Spark 2.2

How was this patch tested?

Add a test suite into DataFrameSuite

[SparkQA]

Test build #77419 has finished for PR 18119 at commit e39bda0.

• This patch passes all tests.
• This patch merges cleanly.
• This patch adds the following public classes (experimental):
• case class InterpretedPredicate(expression: Expression) extends BasePredicate

[kiszk]

ping @zsxwing

[marmbrus]

Since this is purely a fallback for a case that would error, it seems possibly okay to include in 2.2. @zsxwing could you take a look and make sure I'm right?

[zsxwing]

@kiszk I just pushed #17087 directly to branch-2.2 since there is no conflict. Could you close this one?

In addition, I noticed that the test outputs too many logs, could you submit a PR to use testQuietly instead?

[kiszk]

@zsxwing , sure, I submit another PR.

[dongjoon-hyun]

Thank you so much!