pkg/util/metric: fix HistogramMode for streamingingest histograms

Ref: cockroachdb#96514

The above patch migrated all histogram constructors over to
use a generic constructor, which has the ability to specify
a HistogramMode. While preparing a backport of this patch,
I noticed that the following streamingest histograms were
incorrectly tagged with a HistogramMode of
HistogramModePreferHdrLatency:

1. FlushHistNanos
2. CommitLatency
3. AdmitLatency

If you look at the original migration, when HDR was still being
used, these histograms were *not* using the HDR latency defaults:

cockroachdb@a82aa82#diff-1bc5bdba63149e8efeadce17e7eb62bb5cd1dcee22974b37881a627e13c0501dL137-L143

This patch fixes these histograms to no longer incorrectly specify
the HistogramModePreferHdrLatency mode in the histogram options.

This was also fixed in the backport PR.

Release note: none

pkg/ccl/streamingccl/streamingest/metrics.go

@@ -155,23 +155,20 @@ func MakeMetrics(histogramWindow time.Duration) metric.Struct {
 			Buckets:  metric.BatchProcessLatencyBuckets,
 			MaxVal:   streamingFlushHistMaxLatency.Nanoseconds(),
 			SigFigs:  1,
-			Mode:     metric.HistogramModePreferHdrLatency,
 		}),
 		CommitLatency: metric.NewHistogram(metric.HistogramOptions{
 			Metadata: metaReplicationCommitLatency,
 			Duration: histogramWindow,
 			Buckets:  metric.BatchProcessLatencyBuckets,
 			MaxVal:   streamingCommitLatencyMaxValue.Nanoseconds(),
 			SigFigs:  1,
-			Mode:     metric.HistogramModePreferHdrLatency,
 		}),
 		AdmitLatency: metric.NewHistogram(metric.HistogramOptions{
 			Metadata: metaReplicationAdmitLatency,
 			Duration: histogramWindow,
 			Buckets:  metric.BatchProcessLatencyBuckets,
 			MaxVal:   streamingAdmitLatencyMaxValue.Nanoseconds(),
 			SigFigs:  1,
-			Mode:     metric.HistogramModePreferHdrLatency,
 		}),
 		RunningCount:                metric.NewGauge(metaStreamsRunning),
 		EarliestDataCheckpointSpan:  metric.NewGauge(metaEarliestDataCheckpointSpan),