Add min, median and max columns to AccumProfileResults

[nartal1]

This fixes #521

This PR calculates and adds min, median and max columns to sql_plan_metrics_for_application.csv output file. These parameters helps in analyzing which Exec is taking more time or if the median data size of any particular task is large which could slow down the entire stage.

Sample Output:

appIndex,sqlID,nodeID,nodeName,accumulatorId,name,min,median,max,total,metricType,stageIds
1,7,0,"Execute InsertIntoHadoopFsRelationCommand",0,"number of written files",0,0,0,2,"sum","1"
1,7,0,"Execute InsertIntoHadoopFsRelationCommand",1,"written output",0,0,0,751,"size","1"
1,7,1,"Exchange",6,"data size",0,16,16,16,"size","0,1"
1,7,1,"Exchange",7,"number of partitions",0,0,0,320,"sum","0,1"
1,15,5,"HashAggregate",335,"peak memory",0,262144,83886080,113802739712,"size","10"
1,15,5,"HashAggregate",337,"time in aggregation build",0,23612,51375,125070838,"timing","10"
1,15,7,"WholeStageCodegen (3)",340,"duration",0,23618,52598,125466719,"timing","10"
1,15,8,"BroadcastHashJoin",341,"number of output rows",0,0,0,72483658,"sum","10"
1,15,10,"Filter",342,"number of output rows",0,0,0,59007195589,"sum","10"
1,15,11,"ColumnarToRow",343,"number of output rows",0,0,0,79911857034,"sum","10"
1,15,11,"ColumnarToRow",344,"number of input batches",0,0,0,19512293,"sum","10"
1,15,3,"Exchange",332,"shuffle write time",0,1164323,804158883,5600970399,"nsTiming","10,11"

[amahussein]

Thanks @nartal1
I have some few questions since I am not familiar with that part of logic.

[amahussein]

Can we define this as a boolean function? Jus to:

• be able to expand if we want to support other metrics types specific to platform.
• to be consistent to our way of handling ignoredExpressions and types of supported joines.

[amahussein]

Thanks @nartal1