Custom histogram bucket misconfiguration prevents accurate latency percentile tracking

Custom metrics histograms with incorrect bucket ranges fail to accurately track latency percentiles, especially tail latencies (p95, p99)

When creating custom histograms with bentoml.metrics.Histogram(), set buckets argument to cover expected value range. Last bucket must be float('inf'). Number of buckets increases tracking granularity. Example buckets: (0.005, 0.01, 0.025, 0.05, 0.075, 0.1, 0.25, 0.5, 0.75, 1.0, 2.5, 5.0, 7.5, 10.0, float('inf'))