No asyncio global settings

[seemethere]

Overview

This pull request aims to fix issues like #381, #305, #275, and #152. It adds an unambiguous way to get the loop from sanic and works to remove the need to set global asyncio settings as identified by @JendaPlhak here. Which in my opinion third party libraries, like sanic, should not do.

Example:

from sanic.policy import get_event_loop

def thing_that_needs_event_loop():
    loop = get_event_loop()

Implementation

Basic implementation was done by storing an instantiated uvloop policy inside of sanic.policy and then using that to set the global loop for the sanic server at that time.

What this means going forward

1. No more passing the loop

• Server events before/after_start will be the preferred way of sharing the loop but you can also get it through sanic.policy.get_event_loop

2. There is not going to be a way to have a non-blocking way to start a Sanic application

• There is no way to get that functionality to work with multiprocessing
• The use cases could be fulfilled in different ways (like running separate apps that talk to each other through an API)

[r0fls]

I think it would make sense to add this as a property of the sanic app itself. E.g., have sanic.loop return the results of get_event_loop(). What do you think? Also, does this get_event_loop always return the current, running event loop?

[seemethere]

It could possibly be good, maybe each app has it's own instantiated event policy?

[seemethere]

This can be revisited later