Support shared session running mode in Spark Kyuubi &Spark JobServer

[crabo]

Long Running Spark as a SparkSQL query Service or ETL ad-hoc Service is very common nowadays, the key-point is "Shared Session accross jobs" as Spark JobServer or Kyuubi introduced, NO spark sumit any more.

And we can tell too much conf or cache settings are bound to one-time job only, eg:
spark.conf.set("spark.tispark.stale_read", 1651766410000L)
spark.sql("select * from test.t")

lots of "Region missing" and "EpochNotMatch" errors among jobs:
RegionManager.getRegionByKey()
22/05/09 WARN RegionStoreClient: Re-splitting region task due to region error:EpochNotMatch current epoch of region 3291273 is conf_ver: 3659 version: 50821, but you sent conf_ver: 3659 version: 50819

Please take "Shared Session" in account. and Thanks for your great works!

[shiyuhang0]

Thank you for your suggestion, we will investigate the "Shared Session".