admin: Include tokio runtime metrics in the default metrics export

[olix0r]

This change adds a new independent crate, kubert-prometheus-tokio, that exports Tokio runtime metrics when the tokio_unstable compiler feature is enabled. When these unstable features are enabled, a set of tokio runtime metrics is included in the default metrics export. For example:

# HELP tokio_rt_workers The number of worker threads used by the runtime.
# TYPE tokio_rt_workers gauge
tokio_rt_workers 16
# HELP tokio_rt_park Total number of times worker threads parked.
# TYPE tokio_rt_park counter
tokio_rt_park_total 1237
# HELP tokio_rt_noop Number of times workers unparked but found no new work.
# TYPE tokio_rt_noop counter
tokio_rt_noop_total 776
# HELP tokio_rt_steal Number of tasks stolen by workers from others.
# TYPE tokio_rt_steal counter
tokio_rt_steal_total 19
# HELP tokio_rt_steal_operations Number of times workers stole tasks from other.
# TYPE tokio_rt_steal_operations counter
tokio_rt_steal_operations_total 19
# HELP tokio_rt_remote_schedule Total number of remote schedule operations.
# TYPE tokio_rt_remote_schedule counter
tokio_rt_remote_schedule_total 27
# HELP tokio_rt_local_schedule Total number of local schedule operations.
# TYPE tokio_rt_local_schedule counter
tokio_rt_local_schedule_total 546
# HELP tokio_rt_overflow Total number of overflow operations.
# TYPE tokio_rt_overflow counter
tokio_rt_overflow_total 0
# HELP tokio_rt_polls The number of tasks that have been polled across all worker threads.
# TYPE tokio_rt_polls counter
tokio_rt_polls_total 468
# HELP tokio_rt_busy_seconds Total duration of time when worker threads were busy processing tasks.
# TYPE tokio_rt_busy_seconds counter
# UNIT tokio_rt_busy_seconds seconds
tokio_rt_busy_seconds_total 0.08628424000000004
# HELP tokio_rt_injection_queue_depth The number of tasks currently scheduled in the runtime's injection queue.
# TYPE tokio_rt_injection_queue_depth gauge
tokio_rt_injection_queue_depth 0
# HELP tokio_rt_local_queue_depth The total number of tasks currently scheduled in workers' local queues.
# TYPE tokio_rt_local_queue_depth gauge
tokio_rt_local_queue_depth 0
# HELP tokio_rt_budget_forced_yield Number of times a worker thread was forced to yield due to budget exhaustion.
# TYPE tokio_rt_budget_forced_yield counter
tokio_rt_budget_forced_yield_total 0
# HELP tokio_rt_io_driver_ready Number of times the IO driver was woken up.
# TYPE tokio_rt_io_driver_ready counter
tokio_rt_io_driver_ready_total 80

The kubert-prometheus-tokio crate provides standalone prometheus-client integration using the tokio-metrics crate. This is achieved by spawning a background task that updates these metrics once per second. The polling interval for the default metrics endpoint cannot currently be configured.