doc: design for the cluster-controller legacy removal (reference only, do not merge)

[aljoscha]

Not intended to merge. This sits on top of the stack as a reference for what
the code PRs do and why. Read it first, merge it last or never.

The design records the parts that are easy to rediscover incorrectly:

• why the controller and catalog-open builtin-replica reconciliation do not
  conflict
• why the old pending field and its migration remain after runtime pending
  replicas are deleted
• why every managed reshape, including a zero-timeout commit, goes through the
  durable controller record
• why the sequencer cannot safely reproduce the complete strategy decision
  without recreating staged off-loop signal collection
• how the cluster-catalog notification keeps normal zero-timeout latency to one
  controller pass
• why first-pass resource validation applies a signed replacement to live
  inventory and reserves a warranted burst before it materializes
• why create-time scoped configuration belongs in coordinator catalog
  transaction machinery rather than SQL or the cluster controller

In particular, the rejected synchronous sequencer alternative records the bug
found in review: a baseline-only cut-over dropped an active hydration-burst
replica and the next controller tick recreated it cold.

Stack, in merge order

Merged:

1. adapter: remove ENABLE_CLUSTER_CONTROLLER and the legacy REFRESH scheduler #38101 - adapter: remove ENABLE_CLUSTER_CONTROLLER and the legacy REFRESH scheduler
2. adapter: let the cluster controller own system clusters #38102 - adapter: let the cluster controller own system clusters

Open:

3. adapter: delete the staged cluster reconfiguration and direct replica paths #38103 - adapter: delete the staged cluster reconfiguration and direct replica paths
4. sql: accept WITH (WAIT ...) on ALTER CLUSTER unconditionally #38104 - sql: accept WITH (WAIT ...) on ALTER CLUSTER unconditionally
5. this reference PR

The GitHub base still has the old 06-compression-fix branch name because the
stack guard refuses retargeting. That branch points at the exact #38104 commit,
so this PR's diff is still only the design document.

[aljoscha]

Closing to move this off the obsolete 06-compression-fix base alias. GitHub's stacked-PR guard refuses a base change, so the same reference document is being reopened on top of #38104. No review state is lost, this PR had no reviews or comments.