chore(config/tracer): migrate debugAbandonedSpans

[mtoffl01]

What does this PR do?

Tracer migrates to using Config.debugAbandonedSpans instead of its own config.debugAbandonedSpans.

Motivation

https://datadoghq.atlassian.net/browse/APMAPI-1748

Reviewer's Checklist

• Changed code has unit tests for its functionality at or near 100% coverage.
• System-Tests covering this feature have been added and enabled with the va.b.c-dev version tag.
• There is a benchmark for any new code, or changes to existing code.
• If this interacts with the agent in a new way, a system test has been added.
• New code is free of linting errors. You can check this by running ./scripts/lint.sh locally.
• Add an appropriate team label so this PR gets put in the right place for the release notes.
• Non-trivial go.mod changes, e.g. adding new modules, are reviewed by @DataDog/dd-trace-go-guild.

Unsure? Have a question? Request a review!

[pr-commenter]

Benchmarks

Benchmark execution time: 2025-12-05 18:45:54

Comparing candidate commit 44f2aa9 in PR branch mtoff/debugAbandonedSpans with baseline commit 8eb1f41 in branch mtoff/revamp-debug-api.

Found 0 performance improvements and 0 performance regressions! Performance is the same for 3 metrics, 0 unstable metrics.

[hannahkm]

(For here and every other place) is it ever possible for internalConfig on the tracer.config to be nil? Or for config itself to be nil?

[mtoffl01]

Possible, yes, if tracer has not initialized internalConfig via the Get() function, but Config is designed with the expectations that callers must use Get() to get the instance before calling getters/setters.

[mtoffl01]

The alternative is to make every getter/setter on Config check if the receiver c is not nil, but I think it's more standard Golang to design under the assumptions described above.