Use first valid extracted style for distributed tracing

[TonyCTHsu]

What does this PR do?

Currently, when extracting from multiple contexts, our propagation would validate against each other to ensure the trace id and span id are identical. This approach ensure that if user includes Datadog proprietary propagation, those proprietary tags would be propagated and the user will remain having full functionalities that Datadog offers.

If there is a mismatched being detected, since we are not able to identify the source of truth, it would leads to a broken distributed trace. The third scenario on the right illustrated a possible scenario with a mismatched detected given 128 bits trace ID upstream.

This PR changes the propagation behaviour by removing the validation phase for multiple extractions. This means the first successful extraction would always wins. W3C trace context propagation is fully compatible but user prioritizing B3 extraction would lose Datadog's proprietary tags.

System test for trace header precedence: DataDog/system-tests#1454

[marcotc]

This is a change of behavior (users that currently have conflicting propagation headers will see a behavior change IF: they have multiple extraction styles configured and Datadog is one of their extraction styles and Datadog is not the first extraction style).

This is unlikely to happen, but not impossible.

Given that all other traces behave like this PR, taking the extraction value from the first extraction style that is successfully parsed, I don't see much issue with this given. External services with Datadog propagation, written in other languages, are all already using the first extracted style, this then PR increases configuration consistency in heterogeneous deployments.

[marcotc]

Note here: check if there are any system-tests that cover this change (or if the Ruby tracer has been skipped in any relevant system-test because this change wasn't implemented before).

[codecov-commenter]

Codecov Report

Merging #2879 (ad5ad9a) into master (8d3ef4f) will decrease coverage by 0.01%.
The diff coverage is 100.00%.

@@            Coverage Diff             @@
##           master    #2879      +/-   ##
==========================================
- Coverage   98.08%   98.07%   -0.01%     
==========================================
  Files        1301     1301              
  Lines       72065    72058       -7     
  Branches     3310     3307       -3     
==========================================
- Hits        70683    70671      -12     
- Misses       1382     1387       +5     

Impacted Files	Coverage Δ
lib/datadog/tracing/distributed/propagation.rb	91.89% <100.00%> (+0.78%)	⬆️
...ec/datadog/tracing/distributed/propagation_spec.rb	100.00% <100.00%> (ø)

... and 2 files with indirect coverage changes

📣 We’re building smart automated test selection to slash your CI/CD build times. Learn more

[marcotc]

With system-tests coverage, I'm confident this will bring us closer to how Datadog advertises propagation precedence: DataDog/system-tests#1454