refactor(blobs): seperate add node to shard worker

[hannahhoward]

Goals

seperates out the process of adding nodes to shards to its own worker, by renaming and elevating nodes_in_shards to a node_uploads table with its own model -- and the ability to not be in a shard. This is useful cause the add-node-to-shard process is potentially error prone, so it's important to have a records of what shards need to be assigned still to nodes.

Implementation

• new table node_uploads, instead of nodes_in_shards
• pk is cid/space_did/upload_id (will only every assign to one shard_id)
• migration to convert the existing nodes_in_shards to node_uploads
• now, when we create a node, we create a node_upload record to record that it needs to be put in a shard
• we have a query to check nodes not in shards
• we have a step to add all outstanding nodes to shards
• this does mean we have to read it off disk one more time to hash (when the shard is closed), but that's probably ok
• we have a new worker to handle all this.