fix: position of white noise middleware

[pauldambra]

see see https://posthoghelp.zendesk.com/agent/tickets/19227 where user is reporting slow asset load times through the managed reverse proxy

(i believe that) for some routes assets are loaded through a CDN. For others - including the managed reverse proxy - they are loaded by hitting the django web instances. in these cases we see consistently slow time to first byte (TTFB).

---

The whitenoise middleware docs (https://whitenoise.readthedocs.io/en/stable/django.html#enable-whitenoise) say its middleware should be high up in the list of middleware

we had a lot of middleware ahead of it, so we're running those middleware on requests for static assets

locally this change makes the request 60-100ms TTFB instead of 120-169ms TTFB

---

switching on whitenoise's build time compression makes that 30-60ms per request since we're not gzipping the assets on the fly but it is super slow to run the offline compression, so i'm not doing that here

[timgl]

I would definitely put this behind ShortCircuitMiddleware, as otherwise you're loading this for every capture/decide endpoint hit.

[pauldambra]

dropping it below the AllowIPMiddleware didn't seem to have an impact on response time (in my super scientific refreshing locally testing) so went below those two

[benjackwhite]

No more context than anyone else to know if this is a good call or not but in principle - sure?