[processor/tailsampling] record sampling policy

[djluck]

Description

Adds support for optionally recording the policy (and any composite policy) associated with an inclusive tail processor sampling decision:

Link to tracking issue

Fixes
Resolves #35180.

Implementation notes

• This functionality lives behind a feature flag that is disabled by default
• The original issue described a solution where we might attach the attribute solely to the root span. I'm not sure I agree with the commenter that we can rely on this (e.g. we might decide to sample halfway through a long-running trace) so I have attached the attributes to all present scope spans. This feels like a decent trade off between complexity + network cost, as finding the highest non-root parent would require multiple passes of the spans and keeping all span ids in a set

Testing

• Added automated tests to verify enabling the flag both records the expected decision while not impacting existing logic
• Built a custom version and ran it in our preprod environment to ensure it was stable over a 1h period (still evaluating, will update PR with any further observations)

TODO

Does this require a CHANGELOG entry?

[linux-foundation-easycla]

• ❌ The email address for the commit (0fff4e1) is not linked to the GitHub account, preventing the EasyCLA check. Consult this Help Article and GitHub Help to resolve. (To view the commit's email address, add .patch at the end of this PR page's URL.) For further assistance with EasyCLA, please submit a support request ticket.