[SPARK-23907][SQL] Revert regr_* functions entirely

## What changes were proposed in this pull request?
This patch reverts entirely all the regr_* functions added in SPARK-23907. These were added by mgaido91 (and proposed by gatorsmile) to improve compatibility with other database systems, without any actual use cases. However, they are very rarely used, and in Spark there are much better ways to compute these functions, due to Spark's flexibility in exposing real programming APIs.

I'm going through all the APIs added in Spark 2.4 and I think we should revert these. If there are strong enough demands and more use cases, we can add them back in the future pretty easily.

## How was this patch tested?
Reverted test cases also.

Closes #22541 from rxin/SPARK-23907.

Authored-by: Reynold Xin <[email protected]>
Signed-off-by: hyukjinkwon <[email protected]>

sql/catalyst/src/main/scala/org/apache/spark/sql/catalyst/analysis/FunctionRegistry.scala

@@ -300,15 +300,6 @@ object FunctionRegistry {
     expression[CollectList]("collect_list"),
     expression[CollectSet]("collect_set"),
     expression[CountMinSketchAgg]("count_min_sketch"),
-    expression[RegrCount]("regr_count"),
-    expression[RegrSXX]("regr_sxx"),
-    expression[RegrSYY]("regr_syy"),
-    expression[RegrAvgX]("regr_avgx"),
-    expression[RegrAvgY]("regr_avgy"),
-    expression[RegrSXY]("regr_sxy"),
-    expression[RegrSlope]("regr_slope"),
-    expression[RegrR2]("regr_r2"),
-    expression[RegrIntercept]("regr_intercept"),
 
     // string functions
     expression[Ascii]("ascii"),