as qualified identifiers

prevent error for:
!+! a b -> a
!+! a b = a

Now the parser removes the offisde poistions from the stack instead of the scanner.

   case dynamic 1 :: Int of ...

- reordered alternatives of toString (Token) to match the order of the constructors
  in the type definition of Token, for easy comparison

various small things like operators of the form ::>

idents that start with an underscore are only allowed in modules for which
the module name starts with an underscore
Because of this bug fix StdEnv has to be updated (_createArray in the Array class)

renamed UseLayoutBit to ScanOptionUseLayoutBit

(bug_incomplete_instance_def, bug_layout_rule, bug_nested_guard_in_otherwise,
 parse-bug-18, parse_bug_Real_as_class_name, parse_bug_case,
 parse_bug_constructor_with_name_of_basic_type, parse_bug_lost_brackets_in_pattern,
 parse_bug_no_layout_rule)

Tested with compiling Object IO and butstrapping.

bugfix in scanner: reals were not treated correctly

module t
/*2.0
from m import :: T(C1)
0.2*/
//1.3
from m import T, C1
//3.1

will be transformed into

module t
/***/
from m import :: T(C1)
/***/
/*1.3
from m import T, C1
3.1*/

before scanning. In this way we achieve that the upper source can be
compiled with the 2.0 compiler as well as with the 1.3 compiler

The comments _must_ begin the line (no leading whitespaces are allowed).