type RowKey = (String, Vec<(String, String)>, Option<u64>, Option<MetricKind>, bool, String);Expand description
Full identity of one source row: metric name, sorted labels, value, metric kind, name validity, and help.
A null value is its own distinct row identity, not a stand-in for any particular number,
so it is kept apart from every Some(_) identity rather than coerced to
one.
The name and labels lead the tuple so that a BTreeMap<RowKey, _> keeps all rows of one
(metric_name, labels) series adjacent.
metric_kind is the planner’s classification (None for any unsupported metric_type), not
the raw string. Two source rows that differ only in which unsupported type they carry (e.g.
"histogram" vs. "summary") now share one identity instead of two. That only affects the
granularity of the skipped count for rows that are never published either way.