Performance tests are invoked via "sbt perf:test". Currently testing
performance of CEGIS on some random synthesis problems

- NormalizationRule becomes priorities, so that we can have multiple
  distinct layers

- Use the @library annotation, move synthesis stuff to synthesis,
  Oracles.

- Make sure tests use PreprocessingPhase and import synthesis when
  adequate

- Extract proper package objects fix patternRecons and simplifiers

- Reorganize library:
  - leon.{choose,???} -> leon.lang.synthesis
  - leon.{waypoint,epsilon} -> leon.lang.xlang

- Implement the Leon Library in Leon-land rather than Scala-land.
  import leon.Utils._ becomes import leon.lang._
  import leon.Annotations._ becomes import leon.annontation._

  For now, the library defines generic Options and Lists.
  The library is automatically imported from the ./leon script, unless
  the --library=no option is passed.

- Support parsing of multiple files and modules.

- Introduce new annontations:
    @ignore: remove definition from Leon
    @verified: do not consider for verification unless explicitly
        specified

For now only synthesis tests that 100% pass. The rest of benchmarks need
to be investigated due to memory consumption issues