Fixing scan partitioner

[hkaiser]

This is based from a solution as proposed by @sithhell. It fixes the test failures around scan_by_key we're currently seeing.

[sithhell]

Fine by me. It would be really interesting to get the cause of the failure though...

[taeguk]

I can't understand what points of this PR fix the test failures.
Could you explain to me?
It seems that this PR just do only changing order of calling f2 and f3. I want to know what it means. As I think, the order doesn't be meaningful.

hpx/hpx/parallel/util/scan_partitioner.hpp

Lines 129 to 133 in 7fc0361

	finalitems.push_back(dataflow(hpx::launch::sync,
	f3, it, size, prev, curr));
	workitems.push_back(dataflow(hpx::launch::sync,
	f2, prev, curr));

#2325 (just for github link connection.)

[hkaiser]

Yes, you're right. The order should not matter. However, the scan_by_key test breaks if you change the sequence of those calls. We don't know ourselves what's actually happening.