Improve CRLF support in lexer

[jiribenes]

This change should help with resolving #1183, though I can't confirm it's solved completely while being on a Mac.

I also changed the Windows smoke tests to run the case studies -- hopefully this means a slightly bigger subset of the language being checked on Windows CI :)

[jiribenes]

effekt/effekt/shared/src/main/scala/effekt/Lexer.scala

Lines 330 to 335 in e15eca1

	private def advance(): Char =
	val ret = currentChar
	currentChar = nextChar
	nextChar = if charIterator.hasNext then charIterator.next() else '\u0000'
	position = position.advance(ret == '\n')
	ret

is likely also a source of inconsistencies (if it doesn't "magically" work out somehow)

[jiribenes]

Though I don't expect too many CRLF enjoyers using shebangs, we should have a consistent "until the end of the line" predicate, hence this funny-looking change.

[martin-ilgner]

And there are still some tests failing that probably should work, like effekt.ParserTests.Correct multiline span

[jiribenes]

@martin-ilgner if you have the time in the future, could you please try to run the tests again on this branch with an additional patch that changes:

effekt/effekt/jvm/src/test/scala/effekt/ParserTests.scala

Lines 34 to 36 in e15eca1

	def sourceAndPositions: (Source, Seq[Int]) = {
	val lines = content.stripMargin.split("\n").toBuffer
	val positions = scala.collection.mutable.ArrayBuffer[Int]()

into:

def sourceAndPositions: (Source, Seq[Int]) = {
  val normalized = content.stripMargin.replace("\r\n", "\n")
  val lines = normalized.split("\n", -1).toBuffer
  val positions = = scala.collection.mutable.ArrayBuffer[Int]() 
  ...

I looked at the ParserTests and the Position.advance and it feels like the tests are wrong[er]

[martin-ilgner]

@jiribenes I fixed the test (also had to normalize the expected value). I'm still confused by the regex above though.

[jiribenes]

alt. we could use assertNoDiff(source.content, textWithoutSpan)

[jiribenes]

I'm still confused by the regex above though.

WDYM? 🤔
the .split("\n", -1) is Java's "split newlines and keep [trailing] empty lines"