Move Spark 4 usage to shim 0.3.0 and frameless with agnostic encoders

[chris-twiner]

unlike #88 this only applies to Spark 4 / DBR 17, the rest remains.

This largely applies to the SparkHelpers.scala as the implicits from frameless to auto lift codecs should be enough.

• create new testing project, move testutils and ClassicOnly to it
• testutils to self type as suite and provide withFixture to handle ClassicOnly vs swapping out "sparkSession" and running the test twice
• introduce override for conf / gencode handling on connect tests (current behaviour on ClassicOnly tagged tests)
• testShades test runner needs to move to scalatest, move testless' batch runner to the testing, base trait with package name abstract
• pom move from surefire to scalatest plugin (check reports and coverage work)
  migrate other projects to the testing package
• non qualitysparkutils version workarounds split out to specific directories:
  • classicEncoding (pre 4) / agnosticEncoding (4 and above)
  • threeOnly / notThree (ExpressionProxy and Unevaluable foldable = false handling)
  • pre32 / post32 (correct currentSubExprState for FunReWrites)
  • pre33_sparkless / post33_sparkless (collect vs exists on subexpr checks) / no_sparkless (DBRs)