Use SMART resolver strategy when parsing timestamps and dates (apache-spark-on-k8s#504)

This one is really hard to deal with. Since internally we have used `yyyy-MM-dd HH:mm:ss.S` as a default SimpleDateFormat there's no way to keep STRICT parsing and not accidentally cause a failure. This will require rewriting data with any timestamps that have csv

Author: robert3005

FORK.md

@@ -19,6 +19,7 @@
     * yarn: YarnClusterSchedulerBackend, YarnSchedulerBackend
 
 * [SPARK-26626](https://issues.apache.org/jira/browse/SPARK-26626) - Limited the maximum size of repeatedly substituted aliases
+* [SPARK-26178](https://issues.apache.org/jira/browse/SPARK-26178) (partial) - Do not use STRICT strategy when parsing timestamps
 
 # Added
 

sql/catalyst/src/main/scala/org/apache/spark/sql/catalyst/util/DateTimeFormatterHelper.scala

@@ -73,6 +73,6 @@ private object DateTimeFormatterHelper {
       .parseDefaulting(ChronoField.SECOND_OF_MINUTE, 0)
       .toFormatter(locale)
       .withChronology(IsoChronology.INSTANCE)
-      .withResolverStyle(ResolverStyle.STRICT)
+      .withResolverStyle(ResolverStyle.SMART)
   }
 }

sql/core/src/test/scala/org/apache/spark/sql/DateFunctionsSuite.scala

@@ -568,9 +568,10 @@ class DateFunctionsSuite extends QueryTest with SharedSQLContext {
     val y1 = "2016-02-29"
     val y2 = "2017-02-29"
     val ts5 = Timestamp.valueOf("2016-02-29 00:00:00")
+    val ts6 = Timestamp.valueOf("2017-02-28 00:00:00")
     val df2 = Seq(y1, y2).toDF("y")
     checkAnswer(df2.select(unix_timestamp(col("y"), "yyyy-MM-dd")), Seq(
-      Row(ts5.getTime / 1000L), Row(null)))
+      Row(ts5.getTime / 1000L), Row(ts6.getTime / 1000L)))
 
     val now = sql("select unix_timestamp()").collect().head.getLong(0)
     checkAnswer(sql(s"select cast ($now as timestamp)"), Row(new java.util.Date(now * 1000)))
@@ -613,9 +614,10 @@ class DateFunctionsSuite extends QueryTest with SharedSQLContext {
     val y1 = "2016-02-29"
     val y2 = "2017-02-29"
     val ts5 = Timestamp.valueOf("2016-02-29 00:00:00")
+    val ts6 = Timestamp.valueOf("2017-02-28 00:00:00")
     val df2 = Seq(y1, y2).toDF("y")
     checkAnswer(df2.select(unix_timestamp(col("y"), "yyyy-MM-dd")), Seq(
-      Row(ts5.getTime / 1000L), Row(null)))
+      Row(ts5.getTime / 1000L), Row(ts6.getTime / 1000L)))
 
     // invalid format
     checkAnswer(df1.selectExpr(s"to_unix_timestamp(x, 'yyyy-MM-dd bb:HH:ss')"), Seq(