chore: Refactor aggregate expression serde

[andygrove]

Which issue does this PR close?

Part of #1345

Rationale for this change

Refactor in preparation for improving type checking and testing for aggregate expressions.

What changes are included in this PR?

Move aggregate expression serde into individual classes

This is mostly just moving code around. There are only two functional changes:

1. Previously, we had match arms such as case max @ Max(child) if minMaxDataTypeSupported(max.dataType) => which meant that if the type was not supported then we would fall through to the case _ arm which would report unsupported Spark aggregate function, which is misleading. We now do the type checks withing the aggregate serde logic and report unsupported data type for the aggregate instead.
2. I made some small changes to the type checks because the existing code checked for NumericType rather than the specific types that we actually support, so I made this more explicit. We do not support FractionalType, for example, and this is a child of NumericType.

How are these changes tested?

Existing tests.

[codecov-commenter]

Codecov Report

Attention: Patch coverage is 68.55670% with 122 lines in your changes missing coverage. Please review.

Project coverage is 39.18%. Comparing base (f09f8af) to head (2f71857).
Report is 248 commits behind head on main.

Files with missing lines	Patch %	Lines
...main/scala/org/apache/comet/serde/aggregates.scala	67.58%	83 Missing and 35 partials ⚠️
.../scala/org/apache/comet/serde/QueryPlanSerde.scala	83.33%	0 Missing and 4 partials ⚠️

Additional details and impacted files

@@              Coverage Diff              @@
##               main    #1380       +/-   ##
=============================================
- Coverage     56.12%   39.18%   -16.95%     
- Complexity      976     2071     +1095     
=============================================
  Files           119      265      +146     
  Lines         11743    60904    +49161     
  Branches       2251    12935    +10684     
=============================================
+ Hits           6591    23866    +17275     
- Misses         4012    32553    +28541     
- Partials       1140     4485     +3345     

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.

[parthchandra]

It's hard to spot the two functional changes made in this PR because of the large amount of code moved. Can you tag the places where the changes were made?

[andygrove]

It's hard to spot the two functional changes made in this PR because of the large amount of code moved. Can you tag the places where the changes were made?

Sure, functional change #1 moved some of the pre-condition checks. For example, for min, we originally had:

case min @ Min(child) if minMaxDataTypeSupported(min.dataType) =>

And now we have:

case _: Min => CometMin

The pre-condition check is now moved to:

object CometMin extends CometAggregateExpressionSerde {
  override def convert(...) {
    if (!AggSerde.minMaxDataTypeSupported(expr.dataType)) {
      withInfo(aggExpr, s"Unsupported data type: ${expr.dataType}")
      return None
    }
  }
}

Functional change #2 tightened up the checks for supported types:

Example for Sum:

Before, we said that we support all numeric types, including fractional types:

private def sumDataTypeSupported(dt: DataType): Boolean = {
     dt match {
       case _: NumericType => true
       case _ => false
     }
   }

After:

def sumDataTypeSupported(dt: DataType): Boolean = {
     dt match {
       case ByteType | ShortType | IntegerType | LongType => true
       case FloatType | DoubleType => true
       case _: DecimalType => true
       case _ => false
     }
   }

[parthchandra]

lgtm.
(thanks for the additional explanation!)