Composite Head Samplers

[PeterF778]

Changes

Migrating from open-telemetry/oteps#250.

This is another approach for introducing Composite (Head) Samplers. The previous one (open-telemetry/oteps#240) proved too large, with some controversial elements.
This OTEP is a split-off from that one, focusing just on one area - new Composite Samplers.

[github-actions]

This PR was marked stale due to lack of activity. It will be closed in 7 days.

[PeterF778]

Added IsAdjustedCountReliable to SamplingIntent, as discussed by the Sampling SIG.

[jmacd]

Notes: In the prototype I worked on, I decomposed the logic described below, for the parent-sampled case(s). My goal was to preserve the existing ParentBased sampler's ability to distinguish local from remote parents, by registering more delegates. The way I was thinking, there would be 3 delegates: one root, one local-parent, and one remote-parent case. The two parent cases would default to a type named ConsistentParentThreshold case which is the behavior described in lines 250-270.

This lets us advise users who formerly wrote something like

ParentBased(root, localSampled=AlwaysOn, remoteSampled=AlwaysOn, localUnsampled=AlwaysOff, remoteUnsampled=AlwaysOff)

with

ConsistentParentBased(root, localParent=ConsistentParentThreshold, remoteParent=ConsistentParentThresohld)

then, in case the user has remote and/or local-specific attributes they want to add, it'll be clear how they do that.

[PeterF778]

Yes, adding attributes for spans with remote parent is an important use case. I had hoped that this could be achieved using ConsistentRuleBased (where we can test for remote parent) and attribute adding wrappers.

What would ConsistentParentThreshold do for root spans (if misused this way)?

[jmacd]

👍