Support fancy indexing in row iterator

[faltet]

It could be handy to support fancy indexing the Row iterators, so that the next could be possible:

for row in t:
    print "-->", row[[2,1]]

where the third and second fields are selected.

While this should be relatively easy to implement, one should check the impact on performance as well.

[faltet]

I've considered implementing this, but unfortunately NumPy doesn't offer a provision for fancy indexing of record elements of recarrays. In addition, doing this in the Row extension would represent a cost similar to doing it in python space, so I think it is better that this sort of niceties would be done by the user and not clutter the extension with code that will be used quite infrequently.

Closing the ticket.