Which problem is this PR solving?

This PR is part 1 in a series of PRs to add exponential histogram support.

Partially Fixes #3324

Short description of the changes

This PR adds the mapping functions that will be used map values to buckets across the range of usable scales. The scales range from -10 to 20 with higher numbers giving higher resolution. These are not user facing, and will be used internally by the ExponentialHistogramAccumulation to maintain the highest resolution given its max size and range of values observed. The histogram will use the ExponentMapping for scales [-10, 0] and the LogarithmMapping for scales [1, 20]. The Accumulation will start out at max resolution (scale 20) and downscale as needed. It's possible to fit the entire floating point range in two buckets at scale -10, which is the minimum size allowed size for the ExponentialHistogramAccumulation.

This code is heavily based on the Golang reference implementation. For other details see:

• the metrics data model spec
• the metrics sdk spec
• the metrics protobufs

For the other PRs in this series see:

• Part 1 - The mapping functions
• Part 2 - The accumulation and aggregator
• Part 3 - Wiring up export

You can see all 3 PRs combined in the original draft PR

Type of change

Please delete options that are not relevant.

• New feature (non-breaking change which adds functionality)

How Has This Been Tested?

Please describe the tests that you ran to verify your changes. Provide instructions so we can reproduce. Please also list any relevant details for your test configuration

• Extensive unit testing
• End to end test with the 3 PRs from this series adding ExponentialHistogram support

Checklist:

• Followed the style guidelines of this project
• Unit tests have been added
• Documentation has been updated