Numeric histograms should be able to document what kind of bucketing they record with.  This could be useful for downstream processing of these histograms, by allowing earlier/more aggressive data validation, or for doing code generation that supports histogram recording.

Since bucketing schemes are generally used in many histograms, it probably makes sense to use a scheme similar to enums, where each scheme is given a name, perhaps in a buckets.xml file, and then those names are used in histograms.xml.

The scheme declarations should closely match what we have in code, e.g. we might have a declaration like:

<standard-buckets name="COUNTS_100" min="1" max="100" buckets="50/>
<linear-buckets name="PERCENTAGE" min="1" max="100" buckets="101"/>
<custom-buckets name="MAGIC_BUCKETS" bounds="1,5,10,20"/>

and usage:

<histogram name="MyHistogram" units="ms" buckets="COUNTS_100">...</>