I'm attaching a .mcd file, here is a copy of the commit log.

- GRPlatform>>#decodeUtf8Character: fails in gst because the result of
"Character codePoint: NNN" can only be placed in a String for 0<=NNN<=127.
 Above 127 it returns a UnicodeCharacter (which needs to be decoded if you
want to generate a string). In GNU Smalltalk I could simply use "(Character
codePoint: NNN) asString", but that would be as unportable of course.  I
believe a portable application would need to use a codec on a ByteArray,
which is what the proposed implementation does.

- GRUtf8CodecTest>>#testCodecUtf8ShortestForm: allow raising an error.  In
fact, I believe the #deny: is not necessary, and the expected behavior
should be to raise an error always.  The test is currently XFAILed on
Squeak, but passes on gst.

- other changes: compare the result of #decode: against a string literal
instead of a Latin-1 string. Before comparing the result of #encode:,
convert to ByteArray (gst doesn't know the encodings, so it tries to
convert the encoded string from UTF-8 to Unicode; this leads to failure)