Parallel replicas are not implemented for `Merge` tables

On the CI Logs cluster,

```
SELECT count() FROM text_log_15470189561651757858 WHERE message LIKE '%test%'
```
100 GB/sec, uses 10 replicas

```
SELECT count() FROM merge('^text_log_15470189561651757858$') WHERE message LIKE '%test%'
```
10 GB/sec, uses only 1 replica.