ENH: pdist option to return squareform/dense array

[jakirkham]

Would be nice if pdist took an optional flag to return the dense square array (equivalent to running squareform) instead of sparse 1D array.

[tylerjereddy]

Considering the relative ease with which once can call squareform on the result, I'm not entirely clear on the advantage of adding the flag.

The condensed matrix can be quite a bit smaller / more efficient and requiring the user to explicitly use another function to expand to a (potentially much larger) redundant distance matrix doesn't seem like that bad of an idea to me, but others may certainly disagree.

I think there have recently been quite a few changes / attempts to clean up the bloated arguments for pdist and cdist as well.