Fix -loops failure with large-data-transfers

[hwangcc23]

The -loops option is broken after merging the pull-request #57 (#57).

Here is a sample transcript to reproduce the problem.
hwangcc23-BU401LG:~$ dd 'if=/dev/zero' 'bs=2000' 'count=1000' | tr '\000' 'a' | xclip -i -quiet -l 1

Expect: xclip should exist right after any requestor (ex: xclip -o) retrieved data.
What I saw: xclip didn't exist.

Break out the while(1) loop once there is one SelectionRequest completed such that dloop can be decreased by one and compared to sleep for fixing the problem.

[hwangcc23]

Fix #59

[mckellyln]

Hi. This works, but I think we should not remove the other check, as in:

	    if (finished) {
	        del_requestor(requestor);
                break;
	    }

	    if (!requestors) {
	        break;
	    }

Maybe tho it doesn't matter.
I originally had this fix:

            if (sloop > 0 && finished)
                break;

	    if (finished) {
	        del_requestor(requestor);
	    }

	    if (!requestors) {
	        break;
	    }

[hwangcc23]

I think it it better to have the outter loop handle sloop/dloop and the inner loop handle SelectionRequest. :)

[hwangcc23]

@astrand The patch is tested by @mckellyln. Could you review the code and merge it? Thanks.