[nullsafety] null safety errors in inlined code

[Gama11]

#7821 is related to this, but it deserves its own issue.

Inlined code currently breaks the package filters of --macro nullSafety() / @:nullSafety, which is a real problem if the inlined code isn't controlled by you. Especially considering that the std lib isn't null safe yet. For instance, this prevents you from using regexes together with null safety on the Python target:

class Main {
	@:nullSafety
	static function main() {
		~/foo/.match("bar");
	}
}

>haxe -main Main -python foo.py
Main.hx:4: characters 3-22 : Null safety: Cannot assign nullable value here.

Of course Python's EReg should be updated, but that's just a band-aid and I expect there's many more such cases, a lot of them in libraries rather than the std lib.

[RealyUniqueName]

The only workaround I can suggest right now is using @:nullSafety(Off) at expression level:

@:nullSafety(Off) ~/foo/.match("bar");

The way compiler assigns position info to inlined expressions is also a problem with debuggers:

class Tools {
  static public inline function traverse(collection:Array<Int>, callback:Int->Void) {
    for(i in collection) {
      doSomething(i);
      callback(i);
      doAnotherThing(i);
    }
  }
}

class CompletelyDifferentModule {
  static function debugMe() {
    // upon reaching this line you cannot step into `traverse`.
    // you just keep smashing "step over" button 
    // in debugger at this line for `4 * (amount of lines in traverse)` times
    Tools.traverse([1, 2, 3, 4], i -> trace(i)); 
    trace('done');
  }
}

[Gama11]

Right, I remember that FD used to inject --no-inline for Flash/Debug builds to make things behave nicely.. :/