kv: allocator should apply all span config changes proactivily

**Describe the problem**

When a span config change is made it should apply immediately. In some cases the cluster waits until the system has scanned through to find the range instead of applying it immediately.

**To Reproduce**
TODO: This has not been tested yet, but reported by a customer

* Create a table with minimal data in it but a number of splits (e.g. 1000)
* Change the replication factor to 5.
* Notice that not all the ranges change to RF=5 immediately, most do, but some take up to 10 minutes to change.

We should first create a test that replicates this behavior.

**Expected behavior**
We should be reactive to span config changes and apply all the changes immediately.

Jira issue: CRDB-38886