BR restore point (PiTR)'s default `--pitr-concurrency` may be too small

## Enhancement

```
--pitr-concurrency uint32 specify the concurrency to restore log. (default 16)
```

meanwhile the benchmark all uses `--pitr-concurrency=128` 🤷 https://docs.pingcap.com/tidb/stable/br-pitr-guide/#performance-capabilities-of-pitr

in a customer _{(TICKET-8018)} we found that the default setting of 16 is unbearably slow (2 files/second), increasing this to 128 brings it back to acceptable speed (30 files/second).