Using system Python 3 distribution (3.7.3) in order to allow the use of Debian packages for numpy, scipy, etc. without the need of building them or installing build dependencies.
Changed `#!/bin/sh` in shell scripts to `#!/bin/bash` to make `set -o pipefail` work.
- Change celery app to not be a Django app, more like a WSGI app
- Define a Celery task in the Django users app
- Write a test to execute the task
- Update scripts to use the new app to start workers
- Update documentation
Fix#865
* Integrate Flower with Docker Compose setup locally
* Remove alien worker celeryd option
* Move Flower COPY section below the worker's
* Remove set -o pipefail command from Flower start script
* Flower client authentication
* Override flower service image name
* Move flower service to the end of local.yml
* Install flower==0.9.2 in all environments
* Introduce production flower service
* Fix local flower start script
* Document Flower integration
* Prettify *.django envs
Rationale: consistency.
* Reference local environment Flower docs from the production's
* 'two more services' -> 'three more services'
* Move to the python:alpine docker image
- Switch the base images for local and production to alpine
- Install extra dependencies for psycopg2, Pillow and cffi
- Change shebang for shell scripts to use sh instead of bash
* Move to the python:alpine docker image
- Migrate group and user creation to Alpine syntax
* Move to the python:alpine docker image
- Remove `function` keyword, unsupported in shell
* Upgrade various places to the latest Python 3.6
* Test support for translations
* Add gettext library, required for translations support
* Add locale folder for translations support with README documenting it
* Update Changelog
* Tweak command to test translations support
* Re-organize compose/ into environment-specific file groups
Closes#1316.
* Commit missing files
That was weird: git failed to commit one specific folder previously
