[couchup]: warn if [replicator] connection_timeout is low on dbs with lots of deletes

## Expected Behavior

Upgrading a database with many deleted docs using [couchup](https://github.com/apache/couchdb/blob/master/rel/overlay/bin/couchup) just works™.

## Current Behavior

When upgrading a database with lots of deleted docs, the default `[replicator] connection_timeout` of 3000 (which is translated to 10000 for _changes requests), won’t be enough to filter through all deleted docs with in 10 seconds.

The result is that replication first slows down and then stalls as _changes requests timeout messages show up in the logs.

## Possible Solution

couchup could detect the low timeout and big deleted document count and suggest raising the config value (or even offer a way to do it from the cli right away), say `couchup --raise-that-limit` e.g.

Start with rel/overlay/bin/couchup and trace how database upgrades use the [replicator] connection_timeout and _changes requests. Determine how deleted-document volume and the configured timeout should be surfaced; done means couchup provides a clear warning or documented way to raise the limit.

Written by the indexing model from the issue text.