Incorrect detection of BOM in text with little endian UTF32 encoding

[mtillmann0]

There seems to be a bug in the BOM detection for the UTF32 little endian code path:
https://github.com/antlr/antlr3/blob/master/runtime/C/src/antlr3filestream.c#L235

Note that nextChar+1 is compared three times instead of +1, +2 and +3.