Closes#1054, Fixes#1002
The problem is if fread() is called multiple times, then
UnicodeConvertor->convert() is called multiple times, which causes
m_pNewBuf to point to the last read in chunk. Then after the entire file
was loaded, getEOLFormatForm(UnicodeConvertor.getNewBuf(), ...) was being
used which was only trying to detect the EOL mode from the last read in
chunk. If this last chunk started with \n then the file was detected as
Unix line endings. The file linked from issue #1002 happened to have just
the right situation where this occurred.