Fix reversed sorting by multiple index + limit

[ale-rt]

Fix reversed sorting by multiple index by forcing the _sort_iterate_resultset sorting method when we have more than one sorting index

Fixes #108

[ale-rt]

This is probably a fix for #81 as well, even if I see that there is more advanced work happening in PR #82

[ale-rt]

Without this patch the added test fails like:

AssertionError: Lists differ: [97, 76, 52, 39, 29, 26, 25, 24, 17, 0] != [99, 98, 97, 96, 95, 94, 93, 92, 91, 90]

First differing element 0:
97
99

- [97, 76, 52, 39, 29, 26, 25, 24, 17, 0]
+ [99, 98, 97, 96, 95, 94, 93, 92, 91, 90]

[dataflake]

I don't have any knowledge about the inner workings of ZCatalog, so I hope Dieter has better input. The unit test appears to show that your changes work, that's all I can go by.

[d-maurer]

The change is obviously safe - but it may disregard optimization potential: reverse sorting can be optimized in the same way as forward sorting and multi-index sorting can be optimized recursively.

[ale-rt]

I would merge this because it fixes an issue observed by many people, even if it might cost some CPU cycles.
I believe it is better to have something that works slow rather than something that (in some case) doesn't work but it is faster.

I see #82 looks promising performance wise, but it seems to me it requires still some work before it gets merged.

Should I backport this patch to the 5.x branch?

[ale-rt]

Thanks for the reviews!

[dataflake]

Should I backport this patch to the 5.x branch?

Unless there's someone is specifically asking for it I wouldn't bother.

[agitator]

@ale-rt backport for 5.x would be great, since we're using that 5.x for Plone 5.2.x

[ale-rt]

@agitator yeah I know, anyway I think we can even switch to latest master, I do not see it so hard