Xinference time-to-first-token histogram

Declared in xinference/core/metrics.py as a module-level aioprometheus Histogram with buckets=(0.05, 0.1, 0.25, 0.5, 1, 2.5, 5, 10, 30, float('inf')). The float('inf') tail bound is NOT recorded as a bucket number here because it is not a finite bound; it is the standard Prometheus overflow bucket and every observation above 30 s lands in it. This is one of only TWO per-request latency histograms in the whole registry - there is no ITL/TPOT histogram and no decode-time histogram, so a TTFT-vs-ITL regression cannot be decomposed here. Xinference main (xorbitsai/inference), read 2026-10-05. NOT ON BY DEFAULT on the worker: the /metrics endpoint is only mounted when XINFERENCE_DISABLE_METRICS is unset AND a metrics exporter host or port is supplied; otherwise launch_metrics_export_server() is never called and the endpoint does not exist. Supervisor-side metrics are deregistered from the worker's registry on purpose (launch_metrics_export_server removes _SUPERVISOR_ONLY_METRICS), so the worker endpoint carries only worker metrics and the supervisor endpoint only supervisor metrics. aioprometheus, not prometheus_client: counters do not auto-suffix _total the way prometheus_client does, so two of these names end in _counter and three do not - match on the literal name.