Log one line per runner as its boot-verify canary finishes

What does this MR do?

Logs one line per runner when its canary finishes, with the enabled opt-in probes and that runner's duration:

{"msg":"Boot-verify runner finished","status":"passed","phase":"boot-verify","runner":"snM1EtxoC","runner_name":"gitlab-runner-p-m","system_id":"s_1b7ff6e2b6c4","checks":["auth","server_url","commands"],"duration_s":41.1}

checks lists the probes by config key in execution order: auth, helper_version, server_url, cache, commands. On failure the same line logs at error level with status=failed and the error. The build output, previously embedded in that error and so repeated by every line that logged it, is now logged once on its own at the point of failure. The phase-level Running boot-verify canary and Boot-verify canary complete lines are unchanged.

Why was this MR needed?

The canary now has five opt-in probes across three MRs, and a shard's log said nothing about which of them were part of a passing run. Reading the config back from the pod was the only way to know whether a green canary had covered auth or the server URL.

Per-check durations are left out. The wall time is dominated by the executor warm-up and the job, and the job has its own logs.

What are the relevant issue numbers?

• gitlab-com/gl-infra/ci-runners&29 (closed)

Follows !7214 (merged), !7265 (merged), !7279 (merged).