DEV-1125 - run duplicate holdings cleanup under sidekiq

[aelkiss]

Motivation

Running this as a single thread was way too slow and crashed in the middle, giving us no easy way to retry or determine what specific cluster might have failed.

This change:

• Queues batches of clusters to clean up as jobs in sidekiq
• Move logic out of bin & adds a phctl command for cleaning up duplicate holdings (this simplifies loading all the dependencies so the sidekiq stuff works rather than needing to duplicate it in the bin script)
• Adds an optimization to avoid re-saving the cluster's holdings if there are no changes
• Fixes sidekiq-web

Reviewing

@mwarin I think this should be fairly straightforward to re-review. Basically, this is the functionality you already looked at, but adding some functionality to queue small jobs in sidekiq with batches of (by default) 100 clusters at a time.

The sidekiq web update comes from https://github.com/sidekiq/sidekiq/wiki/Monitoring#standalone - we missed adding the rack-session gem before but it was required for a rack update we did earlier but hadn't tested since the tests don't load sidekiq-web.

I ran this locally by adding 5000 clusters from the production database, queuing up the jobs, and watching them run.

[aelkiss]

However -- I see the jobs running, but I'm not convinced it's properly cleaning things up (despite the passing tests); will investigate.

[aelkiss]

I think this is an artifact of how I loaded the test data - I get e.g.

date_received: {"$date"=>"2020-07-21T00:00:00.000Z"}

I'll see if I can do a more proper test in dev...

[coveralls]

coverage: 95.067% (-0.02%) from 95.082%
when pulling 3078966 on DEV-1125-duplicate-holdings-cleanup
into 2bf7c3f on main.

[aelkiss]

After loading the data with mongoimport, I verified that running this in dev does indeed clean up the duplicate holdings.

[mwarin]

Tests pass, code makes sense save for a headscratcher or two, no reason not to APPROVE.

[mwarin]

Looks fishy. Should dedupe_holdings have a body?

...or, as I look closer, this might be a remnant from the previous version, and should be deleted.

[aelkiss]

yeah, needs to be deleted. I inlined the method because I wanted to keep track in remove_duplicate_holdings of how many things had been removed (so that we could avoid saving the cluster if nothing had changed)