In case "skip duplicated chapters" is enabled and the next chapter is from a different scanlator, the initial opened chapter should still be used to remove duplicated chapters.
Otherwise, the original scanlator would get lost and the scanlator of the next chapter would get used, which was only a fallback chapter, because the initial chapters scanlator did not have a next chapter.
It's possible that the queue only contains a failed invalid (does not exist anymore in the database) chapter which then causes graphql errors.
In this case the queue was considered empty which prevented it from getting cleared.
Without being able to clear it, the error could not be fixed.
The local server settings were not shown in case the actual server settings were loading or the request failed.
E.g. in case the "server url" was set to an invalid url, it was not possible to correct it again
Server settings request loading and error state were never handled, thus, in case this request was still loading or failed, instead of showing the proper page state, the missing repo info was shown
In case the manga wasn't loaded immediately, trying to access the trackRecords totalCount caused a TypeError because the initial manga object did not specify it
In case no language had been selected yet and the detected browser language is not supported, the selection was set to undefined instead of showing the default language
To prevent unnecessary re-renders.
The current grid state was stored in the session storage which then triggered an immediate re-render because the just saved snapshot got restored again
The current scroll position was preserved when changing the shown content type (browse, latest, filters) or when changing the filters.
Regression introduced with aab3e25031
* Route to "/" in case path is unknown
* Remove deprecated routes
"sources" and "extensions" were merged into "browse" but the routes were never removed, thus, the paths still matched and opened the components
