recycled query teardown

[jbaxleyiii]

Fix for #718, #667, and probably a lot of other loading issues by removing the current subscription before recycling the observable

I'm going to add a test in the morning for this, but for all those impacted, this will be released tomorrow after the flow PR is released in AC. 👍

TODO:

• Make sure all of the significant new logic is covered by tests
• Rebase your changes on master so that they can be merged easily
• Make sure all tests and linter rules pass
• Update CHANGELOG.md with your change

[helfer]

We should make sure to check that this still does what the recycling was intended for. If I'm not mistaken, the observable might actually be torn down as soon as it has zero subscribers, which is why the recycler adds a subscriber of its own. Do we have tests for the recycling behavior?

[jbaxleyiii]

@helfer we do have tests, but I'm going to work to improve them some. I think this change isn't the full fix we want and may require apollo-client changes. With this change, on remount you get a loading state then the data. Ideally it shouldn't pass loading state.

One question for thought though:
If you contents from your query, is the query loading? If data means no loading, then we could return loading unless we have the data in which case we could set loading to false

[jbaxleyiii]

This will fail until apollo-client PR is approved and released

[munyirik]

@jbaxleyiii could you point me to the commit or PR in apollo-client that coincided with this fix? I'm running into #667 on an apollo-client fork. I'm hoping I can apply the same fix.