Split integrations table to clusterwide and cell local tables

As part of Set sharding keys for feature category `integra... (#463856 - closed) work we identified that the current integrations table can't have a sharding_key set for the Cells support, because integrations can be both instance wide and group/project wide. This means an integration can have NULL values in namespace_id & project_id.

As per the !159403 (comment 2002687899) suggestion we should split integrations table into:

1. Clusterwide table (e.g. instance_integrations) where there is no group/project association
2. Cell local table (e.g. integrations) where there is group/project association

This work blocks updating the following tables:

1. db/docs/issue_tracker_data.yml
2. db/docs/jira_tracker_data.yml
3. db/docs/slack_integrations.yml
4. db/docs/zentao_tracker_data.yml
5. db/docs/slack_integrations_scopes.yml

Because these tables depend on integrations table.