Implement the shiny, new REST API v1

[CrsiX]

This issue is used to keep track of all the different tasks needed to be solved for a working and stable API. Some of those tasks may be delayed back or completely dropped -- it's also a raw collection of ideas. If any of those tasks seems to grow in size too heavily, it could be split out into its own issue as well.

General

• Mark the API as version v1 as soon as it's finished
• Add a configuration for the MateBot core API project
• Add a config schema module which can be used to validate the JSON config (Add a JSON schema file for the config file #75)
• Add offline CSS/JS bundles for the docs UI
• Add an icon for the MateBot
• Add an favicon for the MateBot which is used in the swagger / redoc API overview pages (e.g. /docs)
• Get a list of all locations where versions are somehow mentioned to create consistency among them

Assertions

• Ensure that redirects and failures take precedence over the evaluation of preconditions in conditional requests (see RFC 7233)
• After successful DELETE requests, return code 204 (and no ETag header!)
• Every critical database commit() action should be in a try-except block since error handling would not be possible otherwise!
• Make sure that the details field of the APIError schema always contains JSON-parsable content (not only string, but also some better information, e.g. always some kind of JSON-formatted object, but no lists!)
• Ensure that the CHECK constraint for vote values works on the target database, too! Add a unit test for that.
• What happens when a user should be deleted? Cascade (delete) all the transactions, votes, polls, ..., where it was involved? That shouldn't be the case. Should the community user take over all old transactions? => No, users can't get deleted, they will just be permanently disabled.

Features

• Add consumables to the updates schema
• Add support for consumables, using a simple 'GET /consumables' or something, which contains all necessary data
• Use pydantic's from_orm to fill in schema values, instead of the schema property on the model classes #98
• Add a simple way to consume goods (i.e. make transactions for consumables as easy as possible)
• Implement the healthy metric for 'GET /status'
• API callbacks are a fancy way to avoid polling: https://fastapi.tiangolo.com/advanced/openapi-callbacks/
• Re-work the /communisms endpoint (currently ugly)
• Re-work the /refunds endpoint (currently ugly)
• Add active flag, "closed" timestamp and result to ballot model & schema!
• Make use of the User.active field
• Add support for access logs in extra log files via uvicorn configuration (uvicorn.config.LOGGING_CONFIG)
• Add fancy new callback API option instead of polling for clients that provide a HTTP server as well
• Add some kind of per-application key-value storage which can be accessed via API easily (this should ideally make any client-side data storage superfluous)
• Fix/improve/remove weird logging of sqlalchemy
• Fix problems with 'GET /users' since deprecated data in DB creates ValidationErrors
• Management of external users
• Add handling of the "special user" in various applications
• Add an "easy" way to tell if some API endpoint changed during different versions (automatically, without looking into changelog files)
• Re-build the helper functionality to be cleaner
  • Look through all "old" code to cleanup deprecated functionality
  • Make use of improved helper and hook functions
• Add valid & correct application shutdown
  • Close the session of the class Callback in helpers module
  • Add a "shutdown" message to logging
• Rename 'Ballot' to 'Poll' in all cases

Feature requests

• Add a new config endpoint
• Add errno or some kind of unique, unambiguous declaration of the error to the APIError scheme
• Add error: False to all return schemas, except for the APIError schema, where this is actually True
• Add additional currencies with conversions to the internal "amount"
• Add option how to calculate ETag header values (see commit 9fa5ebb for difference!)
• Add some kind of "power" level to the user model and as requirement to some actions, e.g. vouching for someone else or taking part in a ballot etc. -- this could be used to overwork "external" users, too
• Add a versioned schema definition (using a sub-package for version v1 of the API) -- so that a v2 may use any of the "old" schemas, but may also replace, delete or add some
• Allow users to join more than once to a given communism (maybe even adding a completely new feature for this?)
• Add a read-only endpoint (including the required database tables) for multi-transactions (like communisms, but more general)

Debugging

• Debug problems with my new shiny error handler APIException.handle and RequestValidationError
• Fix logging problems of uvicorn and allow to see the colorized logs in the console again
• Fix Validation Errors upon GET transactions, ballots and votes
• Fix problems of ValidationErrors in APIError schema when an unknown HTTP method (not something like e.g. GET or POST has been used)
• Known limitation of ETag handling: if multiple If-Match / If-None-Match headers are set, only the first will be used

Docs, QA & unit tests

• Add other databases to the CI pipeline checks as well (not only sqlite3 backend)
• Create a collection (for docs) of GET endpoints to call to update the WHOLE local cache/state
• Add some more fractions to test
  • Deleting a user
  • Deleting a vote (which may work, depending on the restricted flag)
  • Deleting a ballot (which shouldn't work)
• Fix race conditions in unittests (e.g. test_users, where a cached version of the user model is PUT to the server but may be rejected for being just one second too old)

Endpoints

• Implement /aliases completely
• Implement /applications completely
• Implement /ballots completely
• Implement /callbacks completely
• Implement /communisms completely
• Implement /consumables completely
• Implement /refunds completely
• Implement /transactions completely
• Implement /users completely
• Implement /votes completely
• Implement /settings completely
• Implement /status completely
• Implement /updates completely

Callbacks

• Add the callback router definition
• Add triggering the callback mechanics upon certain steps in the API, namely CREATE, UPDATE or DELETE
• Add unittests for the callbacks as well

Packaging

• Create a setup.py file
• Add the package creation to the CI pipeline
• Make an extra section in the requirements.txt file or somewhere else to allow optional 'ujson' dependency with "pip install matebot_core[extras]" later