changefeedccl: impact of many changefeeds on cluster

As a follow up to the cdc pre-mortem: https://docs.google.com/document/d/1X2D64G_Jdb9S4MxrBYPH_fwKAP3dkUrhc7yUgZZNZAY/edit#

We're worried about CDC performance when running 100, 1k, 10k, …. or more changefeeds in a cluster. What is a reasonable impact? Will the job system work? How about the console? Will debug.zip capturing work? What about monitoring? Is this a supported use case? If not, what’s the limit? Having frontier processors running (many) on one node is going to cause problems we expect.

We should investigate and address the case of 10xing the number of changefeeds on a cluster, from 50->500.

Jira issue: CRDB-26976

Epic CRDB-9398