fix(tracing): parentId refcount underflow

[morrisonlevi]

Description

I was looking at customer crashes and codex was able to symbolize a crash that our crash tracker wasn't, and it looks like a corrupted $parent and crashed when inheriting the service from it.

I'm not sure this test is precisely the same issue, but nonetheless, it crashes.

Reviewer checklist

• Test coverage seems ok.
• Appropriate labels assigned.

[datadog-datadog-prod-us1-2]

🎉 All green!

❄️ No new flaky tests detected
🧪 All tests passed

🎯 Code Coverage (details)
• Patch Coverage: 100.00%
• Overall Coverage: 60.68% (-0.01%)

_{This comment will be updated automatically if new data arrives.
🔗 Commit SHA: a51ed01 | Docs | Datadog PR Page | Give us feedback!}

[pr-commenter]

Benchmarks [ tracer ]

Benchmark execution time: 2026-05-01 02:59:42

Comparing candidate commit a51ed01 in PR branch levi/parent-refcount with baseline commit 82abdf3 in branch master.

Found 3 performance improvements and 4 performance regressions! Performance is the same for 187 metrics, 0 unstable metrics.

scenario:EmptyFileBench/benchEmptyFileDdprof-opcache

• 🟥 execution_time [+101.842µs; +306.038µs] or [+2.794%; +8.395%]

scenario:PDOBench/benchPDOOverhead

• 🟥 execution_time [+7.965µs; +9.772µs] or [+3.302%; +4.051%]

scenario:PDOBench/benchPDOOverhead-opcache

• 🟩 execution_time [-9.987µs; -8.420µs] or [-4.015%; -3.385%]

scenario:PDOBench/benchPDOOverheadWithDBM

• 🟥 execution_time [+7.378µs; +10.072µs] or [+3.062%; +4.180%]

scenario:PDOBench/benchPDOOverheadWithDBM-opcache

• 🟩 execution_time [-9.172µs; -7.808µs] or [-3.695%; -3.145%]

scenario:PHPRedisBench/benchRedisOverhead

• 🟥 execution_time [+28.512µs; +68.767µs] or [+3.143%; +7.582%]

scenario:PHPRedisBench/benchRedisOverhead-opcache

• 🟩 execution_time [-48.743µs; -37.960µs] or [-4.748%; -3.698%]