Requested by @yuzefovich in #110947.

Is your feature request related to a problem? Please describe.

The code in the following watchers is very similar:

• tenant settings watcher
• tenant info watcher
• tenant capabilities watcher
• settings watcher

It is very redundant and yet needs to do things in a precise order to avoid race conditions and deadlocks.
We need to reduce this complexity to increase maintainability.

Describe the solution you'd like

Abstract at least 2 (preferably at least 3) of the watchers into a common library used in multiple places.

Also merge the tenant capabilities and info watchers into just 1.

Epic: CRDB-26691

Jira issue: CRDB-31697