Figure out unique IDs from CRSP

[davclark]

We should figure out if we can legally distribute those IDs.

[davclark]

At this point, a lot of the logic for this is in dlab-finance, but we need to have an architecture for reading in the CRSP data and matching it with the TAQ data as we stream it in. @davclark and @max-eddy should sit down together and work on this, but it'd be great for @max-eddy to take an initial stab. Recommendations:

• Create a separate file in taq/ for reading / cleaning CRSP data
• Add functionality that uses that file from raw_taq to add PERMNO to the TAQ data
• Figure out how to make pandas a soft dependency. As it stands, we could have the above step involve a python dictionary.