[CON-281] - Add CNodeToSpIdMapManager init in stateMachineManager init

[vicky-g]

Description

On deploy, the cadence of which bull adds jobs does not change. Therefore, even though the logic is correct, the job wouldnt run immediately on deploy. This was thus causing the class to not have the proper data to perform reconfig and propagating false negatives that there are mismatched spIDs on staging.

This PR is to always init the class on stateMachineManager init

Tests

Deployed to stage content node 8 with no change in fetch cnode to spID mapping and no more mismatch spIDs occurred following the deploy.

Monitoring - How will this change be monitored? Are there sufficient logs / alerts?

look for the prefix _findReplicaSetUpdatesForUser(): and mismatched spIDlog

[dmanjunath]

@vicky-g so your theory is there's a race condition here where it tries to find reconfigs before the mapping exists? this fix makes sense in that case

[dmanjunath]

this looks good to me but will let @theoilie give the final approval

[vicky-g]

@vicky-g so your theory is there's a race condition here where it tries to find reconfigs before the mapping exists? this fix makes sense in that case

not a race condition, but when the content node is deployed, the bull queue doesnt initialize by adding a job and instead maintains the current cadence. so the existing class data just gets blasted until the next update cnode to sp id mapping happens (on staging, within <= 4hours, since the job runs every 4 hours)

[theoilie]

makes sense! now I can close #3551 since it fixes the same test