command: Log cumulative time spent waiting for spawn tokens per request

We have a threshold-based spawn token log, but it's still difficult to tell on a per-request basis, how much spawn token delays contributed to the overall request time.

This patch introduces a new field command.spawn_token_wait_ms, which tells us on a per-request basis, how much of that request was spent waiting for spawn tokens.

A recent case was the investigation of https://gitlab.com/gitlab-com/gl-infra/production/-/issues/8350.