Use Java char instead of Java short to pass unsigned WORD values to C functions via JNA

[klehmann]

Together with core dev we tracked down an issue in our large text list support in 12.0.2 where we had to use a Java char instead of a Java short to work with unsigned C WORD values, here the textSize passed to ListAddEntry2Ext:

int i=0;
for (String currStr : strList) {
    Memory currStrMem = NotesStringUtils.toLMBCS(currStr, false);
    if (currStrMem.size() > 65535) {
        throw new DominoException(MessageFormat.format("List item at position {0} exceeds max lengths of 65535 bytes", i));
    }

    //somehow these two lines produce different results for the ListAddEntry2Ext call with text lengths >32767 bytes
    //leading to an error "Insufficient memory" in macOS when using a short

    //short textSize = (short) (currStrMem==null ? 0 : (currStrMem.size() & 0xffff));
    char textSize = (char) currStrMem.size();

    short addResult = capi1201.ListAddEntry2Ext(hList,
        false,
        retListSize,
        (short) (i & 0xffff),
        currStrMem,
        textSize,
        true);
    NotesErrorUtils.checkResult(addResult);

    i++;
}

ListAddEntry2Ext is declared like this:

STATUS far PASCAL ListAddEntry2Ext(MEMHANDLE mhList,
                                BOOL fPrefixDataType,
                                DWORD far *pListSize,
                                WORD EntryNumber,
                                const char far *Text,
                                WORD TextSize,
                                BOOL bAllowLarge);

To avoid similar issues we should replace all Java shorts in the JNX codebase that are used as unsigned WORD values with a Java char.