Tamland runner shard saturates on CPU

This was discovered in the context of production#18361 (comment 2029753950).

The tamland runner shard which currently consists of a single node with 4 CPUs is getting CPU saturated whenever we run a tamland job.

This suggests:

• We may be able to get faster job execution by adding capacity.

• If we have concurrent tamland jobs, they will be competing for host CPU.

Concurrency goes up to 20 jobs:

iwiedler@runners-manager-tamland-green-1.c.gitlab-ci-155816.internal:~$ sudo grep ^concurrent /etc/gitlab-runner/config.toml

concurrent = 20

Would giving this machine more CPUs help?