Fix 2.13 collections methods for NodeSeq

[ashawley]

Add boilerplate from Implementing Custom Collections (Scala 2.13).

Fix #392.

[fanf]

With that patch, I get at some places that overloading error:

Error:(730, 30) overloaded method value flatMap with alternatives:
  (f: scala.xml.Node => scala.collection.IterableOnce[scala.xml.Node])scala.xml.NodeSeq <and>
  [B](f: scala.xml.Node => scala.collection.IterableOnce[B])Seq[B]
 cannot be applied to (scala.xml.Node => net.liftweb.common.Box[com.normation.rudder.services.eventlog.RollbackedEvent])
      entry     <- eventlogs \ "rollbackedEvent"

Where eventlogs has type Node.
It happens only in the following kind of code:

  for{
      event     <- xml
      eventlogs <- event.child
      entry     <- eventlogs \ "rollbackedEvent"
      id        <- (entry \ "id").headOption.map(_.text.toInt) ?~! ("error: id can't be null")
    } yield {
      id
    }

Where ?~! is a method on liftweb Box type, and certainly there is implicit implied of the problem, but I don't understand what/how... But actually the code is not homogenous, and I'm not sure it should have compiled before - at least, it doesn't do what it's documented to do.
And it's easily solved by changing the line like:

      entry     <- (eventlogs \ "rollbackedEvent").iterator // or .headOption

In all case, the patch solve the MUCH bigger problem of NodeSeq vs Seq[Node], and it's a relief for us.

Summary: LGTM

Thanks!

[ashawley]

Thanks for trying it out and for this feedback.

In truth, there may not be a lot of tests around for-comprehensions or flatMap in the test suite. Even if your issue was unique to liftweb, these changes probably warrant further study.

[fanf]

Could we have at least a SNAPSHOT release (sorry if I missed them if they exist)? We realy depend on that fix to have a "non totally broken" app.
If it's impossible to get a nightly, I will need to localy publish a scala-xml version with the patch.

[SethTisue]

@fanf a release attempt is underway at #401