Infrastructure to use UTF-EBCDIC rather than UTF-8 as the internal

Infrastructure to use UTF-EBCDIC rather than UTF-8 as the internal
encoding on EBCDIC platforms. This has property that U+0000..U+009F i.e.
a superset of ASCII are invariant under the encoding. This is EBCDIC
friendly as an encoded string can be looked at as being EBCDIC by lexer
sprintf("%d",...) etc. in same manner that a UTF-8 string be considered
ASCII on ASCII machines.

 - re-arrange utf8.h to get ASCII specific vs Unicode generic bits
   seperate.
 - Add some more macros to comprehend different shift amounts and
   possible swizzle in UTF-EBCDIC vs UTF-8. Change utf8.c to use them.
 - add utfebcdic.h which provides UTF-EBCDIC versions of the macros,
   and conditionally #include it.

EBCDIC build as yet untested.  ASCII still fails the one test.

p4raw-id: //depot/perlio@9185