[SPARK-15370][SQL] Fix count bug

# What changes were proposed in this pull request?
This pull request fixes the COUNT bug in the `RewriteCorrelatedScalarSubquery` rule.

After this change, the rule tests the expression at the root of the correlated subquery to determine whether the expression returns `NULL` on empty input. If the expression does not return `NULL`, the rule generates additional logic in the `Project` operator above the rewritten subquery. This additional logic intercepts `NULL` values coming from the outer join and replaces them with the value that the subquery's expression would return on empty input.

This PR takes over https://github.com/apache/spark/pull/13155. It only fixes an issue with `Literal` construction and style issues.  All credits should go frreiss.

# How was this patch tested?
Added regression tests to cover all branches of the updated rule (see changes to `SubquerySuite`).
Ran all existing automated regression tests after merging with latest trunk.

Author: frreiss <frreiss@us.ibm.com>
Author: Herman van Hovell <hvanhovell@databricks.com>

Closes #13629 from hvanhovell/SPARK-15370-cleanup.