Allow MaxScale to also use zstd as a compression option for rebuilding replicas

I did some testing with zstd and it is yielded very fast and seemingly efficient compression results.
On an 8 vCPU server I created a tarball of a 15GB MariaDB directory. Using all 8 vCPUs it took pigz 2 minutes & 45 seconds with a 4.8 GB output. For zstd using all 8 vCPUs it took 45 seconds for 4.7 GB output.
I then created 2 4 vCPU servers. I loaded 8.7GB worth of data to one of the servers. I then did a streaming backup from Server A to Server B. For pigz it took 0:03:53 and averaged a transfer rate of 37.7MiB/s. For zstd it took 0:01:10 and averaged a transfer rate of 125MiB/s. For each command I used mariadb-backup --parallel=2 and told pigz and zstd to use 2 cores.
I ran these commands 4 times each and got similar numbers. Looks like it makes sense to use zstd to speed things up.

I feel like the inclusion of zstd could help improve the performance of the replica rebuild process.