Reduced memory footprint of FeedUpdateInfo objects. Those objects might stay around for a while and accumulate (up to BUFFER_COUNT_BEFORE_INSERT = 20 at the moment), so in order not to fill up the memory it's better to keep as little data as possible.
Previously ChannelInfo data was stored, causing ReadyChannelTabLinkHandler objects to be also stored uselessly (and those channel tabs contain prefetched JSON data which used ~700KB of memory).
The previous "main" tab is now just a normal tab returned in getTabs().
Various part of the code that used to handle channels as ListInfo now either take the first (playable, i.e. with streams) tab (e.g. the ChannelTabPlayQueue), or take all of them combined (e.g. the feed).
Before the Target would sometimes be garbage collected before being called with the loaded channel icon, since Picasso holds weak references to targets. This meant that sometimes a new streams notification would not be shown, because the lambda that should have shown it had already been garbage collected.
It never kicked in since we are never returning a retry() Result, but always either success() or failure() (see createWork() function). Also, there is already a default (exponential backoff starting from 30 seconds), so no need to override it.
Removed some non-translatable strings and just hardcoded them in the code, like it's being done for other string separators. This also deduplicates some code by using Localization.
Used some Kotlin feature to reduce code.
Do not open the setting for a specific notification channel (Settings.ACTION_CHANNEL_NOTIFICATION_SETTINGS), but the settings for all notifications by the app (Settings.ACTION_APP_NOTIFICATION_SETTINGS)