Update deployment process to use conda/pip instead of cloning github repos

[hdoupe]

Last night, I was unable to reset the node environments because it was unable to install btax. In this PR, I refactored the install scripts so that all of the OSPC packages are installed with conda instead of some confusing mix of cloned github repos and pip installations.

All packages are installing correctly now. The only problem is that they are in a different location relative to the PUF file. Thus, if you run TaxBrain on the test app, you get:

taxcalc traceback:

or on CCC you get:

btax traceback:

Traceback (most recent call last):
  File "/home/ubuntu/miniconda2/envs/aei_dropq/lib/python2.7/site-packages/celery/app/trace.py", line 374, in trace_task
    R = retval = fun(*args, **kwargs)
  File "/home/ubuntu/miniconda2/envs/aei_dropq/lib/python2.7/site-packages/celery/app/trace.py", line 629, in __protected_call__
    return self.run(*args, **kwargs)
  File "/home/ubuntu/deploy/taxbrain_server/celery_tasks.py", line 159, in btax_async
    tables = json.loads(runner_json_tables(**user_mods))
  File "/home/ubuntu/miniconda2/envs/aei_dropq/lib/python2.7/site-packages/btax/front_end_util.py", line 172, in runner_json_tables
    out = runner(test_run, start_year, iit_reform, **user_params)
  File "/home/ubuntu/miniconda2/envs/aei_dropq/lib/python2.7/site-packages/btax/execute.py", line 48, in runner
    run_btax(test_run, True, start_year, {}, **econ_params)
  File "/home/ubuntu/miniconda2/envs/aei_dropq/lib/python2.7/site-packages/btax/run_btax.py", line 110, in run_btax
    iit_reform, **user_params)
  File "/home/ubuntu/miniconda2/envs/aei_dropq/lib/python2.7/site-packages/btax/parameters.py", line 207, in get_params
    indiv_rates = get_rates(baseline, start_year, iit_reform)
  File "/home/ubuntu/miniconda2/envs/aei_dropq/lib/python2.7/site-packages/btax/get_taxcalc_rates.py", line 83, in get_rates
    reform=reform)
  File "/home/ubuntu/miniconda2/envs/aei_dropq/lib/python2.7/site-packages/btax/get_taxcalc_rates.py", line 45, in get_calcu

[hdoupe]

Alright, two nodes are up and running on the test app. Progress is slow, but it's happening.

[hdoupe]

Testing strategy:
Detach the worker nodes from the production app one-by-one and update them. If the update was successful, I attached it to the test app which was running on 1.4.0 and checked the log for each each node to make sure that they were all still sending and receiving jobs. I did this for eight of the nine nodes. The ninth is still serving the production app and will be updated when the production app is updated.