Fix MessageUnpacker#unpackString in case that string length is more than...

[komamitsu]

I modified MessageUnpacker#unpackString to fix #209.

@xerial Can you review this change?

[xerial]

I found another bug; This part should use ensure(strLen-cursor):

    while(cursor < strLen) {
                     if (!ensure(strLen))
                         throw new EOFException();

[xerial]

I think there is no guarantee that the read length from the input always matches Math.min(buffer.size() - position, strLen-cursor) when OVERFLOW (output buffer has insufficient space) happens. When OVERFLOW occurs, the input data might be partially read. That is why I compute readLen as bb.limit() - bb.remaining().

[xerial]

Ah. I was wrong. readLen can be retrieved by bb.position() since we created a new ByteBuffer at

ByteBuffer bb = buffer.toByteBuffer(position, readLen);

whose initial position is 0.

[xerial]

Now I got it. I don't need to set readLen in while(bb.hasRamaining()) loop. Thanks