fix OOM kills by reducing worker counts, cuts memory utilization

[irvingpop]

Description of changes

we were using way too much memory and racking up oomkills.

Before:
1 gunicorn master + 3 gunicorn workers + 2 django_q workers = 6 processes per container
At ~65 MB average: ~390 MB per container
Two containers (prod + staging): ~780 MB

After:
1 gunicorn master + 1 gunicorn worker + 1 django_q worker = 3 processes per container
At ~65 MB average: ~195 MB per container
Two containers: ~390 MB (~50% reduction!)

[irvingpop]

confirmed memory use reduction in staging:

Gunicorn:
40.8 MB - gunicorn master
49.5 MB - gunicorn worker
------
90.3 MB subtotal
Django-Q:
27.3 MB - qcluster (monitor/sentinel)
22.3 MB - qcluster (process)
25.2 MB - qcluster (worker - active)
16.8 MB - qcluster (pusher)
16.8 MB - qcluster (guard)
------
108.4 MB subtotal
Total per container: ~199 MB PSS