feat(metrics): measure sink CPU usage - #27097

Summary

Extends the per-component component_cpu_usage_ns_total metric (#25317) to sinks, including the tasks they spawn.

Stacked on #27104 (counter inheritance through a thread-local, spawn_timed / spawn_timed_without_span): review that one first.

Most sink CPU does not run in the sink task: encoding and compression happen in ConcurrentMap tasks (request_builder), and TLS encryption and socket writes in hyper's per-connection tasks. Timing only the sink task would miss most of it.

• Sinks: new measure_cpu_usage option; the sink task is wrapped with cpu_timed and its build() runs under cpu_scoped.
• Subtasks now attributed:
  • ConcurrentMap (request_builder / concurrent_map)
  • hyper client connections and pool tasks, via vector_http::CpuTimedExecutor on the HttpClient pool and the vector sink's Http2Client
  • server connections spawned by crate::http::serve (used by prometheus_exporter; sources have no counter, so they are unaffected)
  • legacy BatchSink service calls
  • tower Buffer worker of distributed services (Buffer::pair instead of BufferLayer)
  • per-sink background tasks: splunk_hec acks, datadog_traces APM stats, redis sentinel repair, websocket_server, prometheus_exporter server, blackhole reporter

Windows accuracy: the measure_cpu_usage doc of sinks carries the same Windows caveat as transforms (see #27104).

Not covered

• librdkafka native threads (kafka sink compression and network I/O)
• tasks spawned internally by the AWS SDK and tonic
• spawn_blocking: the closure runs on a blocking-pool thread, outside any timed poll. Our own call sites (dns.rs, kafka healthcheck) could be wrapped in a follow-up, but two indirect uses cannot be covered, because the spawn_blocking call happens inside the dependency with no hook:
  • tokio::fs writes in the file sink (the write syscall and buffer copy; compression stays in the sink task and is counted)
  • hyper's default DNS resolver for HTTP sinks (one lookup per new connection)

References

Related: #25317
Depends on: #27104

Vector configuration

sinks:
  out:
    type: http
    inputs: [in]
    uri: https://example.com
    encoding:
      codec: json
    measure_cpu_usage: true

How did you test this PR?

• New topology test component_cpu_usage_emitted_sink: the mock completion sink burns CPU in a spawn_timed subtask; the existing no-measure test also checks the sink.
• cargo test on cpu_metrics, topology::test, sinks::util, http, vector-http and the touched sinks/transforms; clippy on the touched feature set.

Does this PR include user facing changes?

• Yes. Please add a changelog fragment based on our guidelines.
• No. A maintainer will apply the no-changelog label to this PR.

Contributor Guidelines

• Please read our Vector contributor resources.
• Do not hesitate to use @vectordotdev/vector to reach out to us regarding this PR.
• Before pushing, follow our pre-push guidance.
• After a review is requested, please avoid force pushes to help us review incrementally.
  • Feel free to push as many commits as you want. They will be squashed into one before merging.
  • For example, you can run git merge origin master and git push.