feat: abortable retry delay

[xpirt]

Context

If a web application implements axios-retry and also a network change detection system, he may want to re-call the API requests even if in the middle of the retries.

Problem

Currently, there is no way to clear the setTimeout and reject the Promise that handles the retries' attempts.

Solution

Implement the possibility to pass an AbortController's signal, so that a developer is able to stop pending retries.

[mindhells]

Hi @xpirt
sorry for the delay.
I have some questions/concerns about this PR:

• changing the prettier rules should be part of a different PR, where you could explain the motivations for the change
• it is important to cover review the tests, and add new ones if needed. Specially when adding new functionality
• would't the axios instance cancellation mechanism work?

[xpirt]

Hi @xpirt sorry for the delay. I have some questions/concerns about this PR:

• changing the prettier rules should be part of a different PR, where you could explain the motivations for the change
• it is important to cover review the tests, and add new ones if needed. Specially when adding new functionality
• would't the axios instance cancellation mechanism work?

1. I see, will revert that. Edit: reverted.
2. Will try to come up with some tests
3. Nope. The axios-retry retry mechanism uses a setTimeout which, despite the request being aborted, it is not cleared properly. This results into the timeout to expire and execute the request afterwards. Let me know if that's clear.

[mindhells]

Hi @xpirt sorry for the delay. I have some questions/concerns about this PR:

• changing the prettier rules should be part of a different PR, where you could explain the motivations for the change
• it is important to cover review the tests, and add new ones if needed. Specially when adding new functionality
• would't the axios instance cancellation mechanism work?

1. I see, will revert that. Edit: reverted.
2. Will try to come up with some tests
3. Nope. The axios-retry retry mechanism uses a setTimeout which, despite the request being aborted, it is not cleared properly. This results into the timeout to expire and execute the request afterwards. Let me know if that's clear.

Is there a way we can honor the original abort signal (if any) from the axios instance instead extending this interceptor interface?

[michal-billtech]

@mindhells @xpirt please take a look at #273. Is this what you had in mind?

[mindhells]

closing this one in favor of #273