Only reset abortcontroller ref if captured instance is same as ref

[EtherZa]

#302 describes a race condition where an aborted request resets the abort controller ref even though a subsequent request has been issued and already replaced the ref.

Code sandbox: https://codesandbox.io/p/sandbox/react-hooks-usefetch-forked-xsdrhl?file=%2Fsrc%2Findex.js

Steps to reproduce:

1. Open sand box
2. View browser network traffic (ctrl + i / Option + ⌘ + I / select network)
3. Repeatedly click "Click Me" before previous request has completed
4. Note that approximately every second request is aborted while the others complete

Cause:
On completing a request (normal or aborted), the captured AbortController ref (controller.current) is set to undefined in the finally block. If a request is aborted and then immediately resubmitted, this results in a race condition where:

1. Request A is started
2. AbortController for request A is captured
3. Request A abort is initiated via abort()
4. Abort signal is sent
5. Request B is started
6. AbortController for request B is captured
7. Request A completes and sets captured controller ref to undefined
8. Request B abort is initiated via abort()
9. Abort signal has been lost by 8 and cannot be issued
10. Request B completes

Resolution:
Only set the controller ref to undefined if the ref is still set to the current controller instance

Note:
This PR makes no attempt to resolve the underlying race condition. It proposes a solution to allow for the last issued request to be aborted while potentially managing the race condition through other means.

[freshollie]

This solves an issue we were experiencing with this library. @alex-cory I'm sure others would benefit from merging