Refactor video uploads (#5570)
* Remove unused video field
* Stop exposing video dispatch
* Move cancellation out of the reducer
* Make useUploadStatusQuery controlled by jobId
* Rename SetStatus to SetProcessing
This action only has one callsite and it's always passing "processing".
* Move jobId into video reducer state
* Make cancellation scoped
* Inline useCompressVideoMutation
* Move processVideo down the file
* Extract getErrorMessage
* useServiceAuthToken -> getServiceAuthToken
* useVideoAgent -> createVideoAgent
* useVideoUploadLimits -> getVideoUploadLimits
* useUploadVideoMutation -> uploadVideo
* Use async/await in processVideo
* Inline onVideoCompressed into processVideo
* Use async/await for uploadVideo
* Factor out error messages
* Guard dispatch with signal
This lets us remove the scattered signal checks around dispatch.
* Move job polling out of RQ
* Handle poll failures
* Remove unnecessary guards
* Slightly more accurate condition
* Move initVideoUri handling out of the hook
* Remove dead argument
It wasn't being used before either.
* Remove unused detailed status
This isn't being used because we're only respecting that state variable when isProcessing=true, but isProcessing is always false during video upload.
If we want to re-add this later, it should really just be derived from the reducer state.
* Harden the video reducer
* Tie all spawned work to a signal
* Preserve asset/media for nicer error state
* Rename actions to match states
* Inline useUploadVideo
This abstraction is getting in the way of some future work.
* Move MIME check to the only place that handles it