[app] deleting source with incomplete message downloads should abort and delete cleanly

[vickiniu]

In the case where we receive a new source with many messages (tested with 5000) but want to actually delete this source before all messages finish downloading, we would expect that:

• Messages queued for download/decryption should be removed from the queue on the pending deletion event
• All messages should eventually be submitted through sync for deletion by the server
• Pending deleted messages should not be re-downloaded in the interim time

Related to #3030

This means that we should:

• Update an item's fetch_progress to be some new state like SCHEDULED_DELETION
• Check all item fetch_progress before incrementing or taking up work + drop
• Only queue N number of items for processing at a time, and enqueue more during sync as needed
• Check that items are not already pending deletion when receiving new sync message

[vickiniu]

Issue seems to be that:

• Not all items for the source have been deleted from the server, and
• There are many pending stale item_seen events being queued for sync. We should detect + purge those