BUG?: gpg doesn't flush attribute stream after writing an attribute

[Adam Milazzo]

It would be very nice if GPG flushed the attribute stream after writing
an attribute, so that the number of bytes reported in the ATTRIBUTE
status message can be read reliably in response to that message.

As it is, the read often blocks because the data is sitting in GPG's
output buffer.

I think all this would require is an fflush() call at the end of
keylist.c:dump_attribs().