Add a lookup function

[johnmyleswhite]

See this example for the semantics of Pandas' lookup() function:

http://stackoverflow.com/questions/13893227/vectorized-look-up-of-values-in-pandas-dataframe

[HarlanH]

Hm, we'd have to do something somewhat different, because we don't have row
indexes. Something like...

lookup(from::DataFrame, index_col::Index, index_val::T1, data_col::Index)

or maybe

lookup(from::DataFrame, row_where::Dict, data_col::Index)

where row_where looks like {"Date" => "2011-02-02 00:00:00"} or even
{"Year" => 2012, "Month" => 5} to index on multiple columns. If we actually
had column or column-set hashes, getting the row could be fast.

[tshort]

They are not pulled in automatically, but there is some code in
src/indexing.jl that can index columns. I haven't checked in a while to see
if it still works. Either way, that code might be a place to look, either
to implement indexing or to generate an index on the fly.

[bkamins]

We now support it via GroupedDataFrame.