[SPARK-35999][SQL] Make from_csv/to_csv to handle day-time intervals properly

### What changes were proposed in this pull request?

This PR fixes an issue that `from_csv/to_csv` doesn't handle day-time intervals properly.
`from_csv` throws exception if day-time interval types are given.
```
spark-sql> select from_csv("interval '1 2:3:4' day to second", "a interval day to second");
21/07/03 04:39:13 ERROR SparkSQLDriver: Failed in [select from_csv("interval '1 2:3:4' day to second", "a interval day to second")]
java.lang.Exception: Unsupported type: interval day to second
 at org.apache.spark.sql.errors.QueryExecutionErrors$.unsupportedTypeError(QueryExecutionErrors.scala:775)
 at org.apache.spark.sql.catalyst.csv.UnivocityParser.makeConverter(UnivocityParser.scala:224)
 at org.apache.spark.sql.catalyst.csv.UnivocityParser.$anonfun$valueConverters$1(UnivocityParser.scala:134)
```

Also, `to_csv` doesn't handle day-time interval types properly though any exception is thrown.
The result of `to_csv` for day-time interval types is not ANSI interval compliant form.

```
spark-sql> select to_csv(named_struct("a", interval '1 2:3:4' day to second));
93784000000
```
The result above should be `INTERVAL '1 02:03:04' DAY TO SECOND`.

### Why are the changes needed?

Bug fix.

### Does this PR introduce _any_ user-facing change?

No.

### How was this patch tested?

New tests.

Closes #33226 from sarutak/csv-dtinterval.

Authored-by: Kousuke Saruta <sarutak@oss.nttdata.com>
Signed-off-by: Wenchen Fan <wenchen@databricks.com>

sql/catalyst/src/main/scala/org/apache/spark/sql/catalyst/csv/UnivocityGenerator.scala

@@ -66,6 +66,11 @@ class UnivocityGenerator(
         IntervalUtils.toYearMonthIntervalString(
           row.getInt(ordinal), IntervalStringStyles.ANSI_STYLE, start, end)
 
+    case DayTimeIntervalType(start, end) =>
+      (row: InternalRow, ordinal: Int) =>
+      IntervalUtils.toDayTimeIntervalString(
+        row.getLong(ordinal), IntervalStringStyles.ANSI_STYLE, start, end)
+
     case udt: UserDefinedType[_] => makeConverter(udt.sqlType)
 
     case dt: DataType =>

sql/catalyst/src/main/scala/org/apache/spark/sql/catalyst/csv/UnivocityParser.scala

@@ -222,6 +222,11 @@ class UnivocityParser(
         Cast(Literal(datum), ym).eval(EmptyRow)
       }
 
+    case dt: DayTimeIntervalType => (d: String) =>
+      nullSafeDatum(d, name, nullable, options) { datum =>
+        Cast(Literal(datum), dt).eval(EmptyRow)
+      }
+
     case udt: UserDefinedType[_] =>
       makeConverter(name, udt.sqlType, nullable)
 

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

@@ -18,7 +18,7 @@
 package org.apache.spark.sql
 
 import java.text.SimpleDateFormat
-import java.time.Period
+import java.time.{Duration, Period}
 import java.util.Locale
 
 import scala.collection.JavaConverters._
@@ -28,6 +28,7 @@ import org.apache.spark.sql.functions._
 import org.apache.spark.sql.internal.SQLConf
 import org.apache.spark.sql.test.SharedSparkSession
 import org.apache.spark.sql.types._
+import org.apache.spark.sql.types.DayTimeIntervalType.{DAY, HOUR, MINUTE, SECOND}
 import org.apache.spark.sql.types.YearMonthIntervalType.{MONTH, YEAR}
 
 class CsvFunctionsSuite extends QueryTest with SharedSparkSession {
@@ -308,4 +309,48 @@ class CsvFunctionsSuite extends QueryTest with SharedSparkSession {
       }
     }
   }
+
+  test("SPARK-35999: Make from_csv/to_csv to handle day-time intervals properly") {
+    val dtDF = Seq(Duration.ofDays(1).plusHours(2).plusMinutes(3).plusSeconds(4)).toDF
+    Seq(
+      (DayTimeIntervalType(), "INTERVAL '1 02:03:04' DAY TO SECOND",
+        Duration.ofDays(1).plusHours(2).plusMinutes(3).plusSeconds(4)),
+      (DayTimeIntervalType(DAY, MINUTE), "INTERVAL '1 02:03' DAY TO MINUTE",
+        Duration.ofDays(1).plusHours(2).plusMinutes(3)),
+      (DayTimeIntervalType(DAY, HOUR), "INTERVAL '1 02' DAY TO HOUR",
+        Duration.ofDays(1).plusHours(2)),
+      (DayTimeIntervalType(DAY), "INTERVAL '1' DAY",
+        Duration.ofDays(1)),
+      (DayTimeIntervalType(HOUR, SECOND), "INTERVAL '26:03:04' HOUR TO SECOND",
+        Duration.ofHours(26).plusMinutes(3).plusSeconds(4)),
+      (DayTimeIntervalType(HOUR, MINUTE), "INTERVAL '26:03' HOUR TO MINUTE",
+        Duration.ofHours(26).plusMinutes(3)),
+      (DayTimeIntervalType(HOUR), "INTERVAL '26' HOUR",
+        Duration.ofHours(26)),
+      (DayTimeIntervalType(MINUTE, SECOND), "INTERVAL '1563:04' MINUTE TO SECOND",
+        Duration.ofMinutes(1563).plusSeconds(4)),
+      (DayTimeIntervalType(MINUTE), "INTERVAL '1563' MINUTE",
+        Duration.ofMinutes(1563)),
+      (DayTimeIntervalType(SECOND), "INTERVAL '93784' SECOND",
+        Duration.ofSeconds(93784))
+    ).foreach { case (toCsvDtype, toCsvExpected, fromCsvExpected) =>
+      val toCsvDF = dtDF.select(to_csv(struct($"value" cast toCsvDtype)) as "csv")
+      checkAnswer(toCsvDF, Row(toCsvExpected))
+
+      DataTypeTestUtils.dayTimeIntervalTypes.foreach { fromCsvDtype =>
+        val fromCsvDF = toCsvDF
+          .select(
+            from_csv(
+              $"csv",
+              StructType(StructField("a", fromCsvDtype) :: Nil),
+              Map.empty[String, String]) as "value")
+          .selectExpr("value.a")
+        if (toCsvDtype == fromCsvDtype) {
+          checkAnswer(fromCsvDF, Row(fromCsvExpected))
+        } else {
+          checkAnswer(fromCsvDF, Row(null))
+        }
+      }
+    }
+  }
 }