Implement zero sentry exception policy for staging

Problem

Staging can, and should be used to catch exceptions, and alert us when the rate of error increases in a short period of time. However, there are many causes of errors (many false herrings, or long standing errors).

In gitlab-com/gl-infra/production#5931 (closed), we had coverage issues. This mean that the error in particular NoMethodError was not caught on CI.

Proposal

Iteration 1

Alert on known errors (using Kibana watches)

Status	Error	Description	Implementation
🚧	NoMethodError for non-nil objects	Objects should never hit NoMethodError. Investigate where the error is coming from by checking the Kibana link for both the pubsub-rails-inf-gstg*, and pubsub-sidekiq-inf-gstg* indexes. Open a new issue, or comment on existing issue there is one.	gitlab-com/runbooks!4134 (merged)
💡	NoMethodError for nil objects

Iteration 2

We assign all other errors by feature category to stage groups - gitlab-com/gl-infra&396 (closed)