Speed up layout of text boxes

[timb07]

Processing some PDFs takes a lot of time in layout.py; in particular (for one PDF I've tried) in LTLayoutContainer.group_textboxes().

This PR has two changes:

1. Remove utils.csort() and replace all uses with native list.sort()
2. In group_textboxes(), make dists a sortedcontainers.SortedListWithKey() instead of a list, which removes the need to re-sort dists repeatedly

Note that replacing csort() (which uses "Decorate-Sort-Undecorate" to ensure the sort is stable) with the native sort doesn't change the behaviour, since Python sorting is stable since Python 2.2.

For the test PDF I have (which motivated these changes), running pdf2txt.py now takes around 7 seconds instead of 17 seconds.

(I also tested these changes with the file in #128, but no speed up was seen for that input.)

[tataganesh]

Why are elements being removed from the dists list? Is it because accumulating elements to be removed, and removing them is less expensive than adding elements to dists (That satisfy the condition if (obj1 in plane and obj2 in plane)?
Also, instead of using the list comprehension, you can loop through the elements in dists, check which among them satisfies the condition, and remove it in the same loop. Something like -

for (c,d,obj1, obj2) in dists:
    if (obj1 in removed or obj2 in removed):
        dists.remove( (c,d,obj1, obj2))

[tataganesh]

For now, I am merging this. Might take a look at the requested changes later.