Remove messagebus shovel

[geropl]

Remove shovel by having ws-manager-bridge "pull" state from each cluster
Goal: Replace messagebus cross-clusters dependency with the (already existing) workspace-cluster dependency.

[geropl]

The basic code change is quite simple: #7523

There are two problems left:

• testing: make sure the updates we're sending are the ones we expect (e.g., same as before)
• fix the "start workspace" case:
  • currently we're broadcasting "artificial" events from the server globally for the preparing phase (this includes initial creation/context parsing/resolution as well as image building!)
  • with removing the "global" option, other regions would not receive updates for workspace starts in a different region until the pending phase (when ws-manager is called and the first pod is created). This is especially important for image builds - because it extends the "time without update" to multiple minutes potentially - , and more so for prebuilds with image builds - which for 50% of our users always happen in a different (US) cluster
  • the ideal solution would be to move image-builder into workspace cluster. But this is non-trivial, has complex dependencies in itself and will not happen soon, so we cannot depend on it
  • alternatively, we can try to emulate the updates for the workspace phase preparing by:
    • having something (in either server or bridge) poll the DB for workspace instances in phase preparing, which are not governed by the current bridge
    • if there is a state change: generate an update and distribute it locally

/cc @AlexTugarev