possible sys_read failure for chunked io

[ostap]

sys_read fails on chunked io when the requested amount of data for read is less than the size of the chunk. we need to implement a buffered io and keep the unread data in the buffer.

here is the problematic code from the system.c:

static int net_read_chunk(IO *io, void *buf, int size)
{
    int r = 0, sz = 0;
    if (readn(io, &sz, sizeof(sz), net_read) == sizeof(sz)) {
        if (sz < 0)
            io->stop |= IO_TERM;
        else if (sz > size)
            io->stop |= IO_ERROR;
        else if (sz > 0)
            r = readn(io, buf, sz, net_read);
    }   

    return io->stop & ~IO_TERM ? -1 : r;
}

the situation with sz > size should put the requested amount of data into the buf and keep the remaining in the buffer associated with the IO.

[ostap]

the issue has been worked around as part of #18 so it is not of a critical priority, however we need to fix it on the IO level at some point.