feat(otel): Add per-PR OTel conformance test suite - #630

Claude AI review

This PR adds a standalone conformance-tests-otel/ module (19 OTel handlers + long-running variants, two SAM templates) and a new otel-conformance-tests.yml workflow. The Java handlers are correct against the current SDK API surface: retry handlers use the 1-based StepContext.getAttempt() with fixedDelay(2, …) correctly, Otel14MapFailure relies correctly on the lenient allCompleted default and throws on !allSucceeded(), Otel8HandledFailure catches the rethrown step exception before the recovery step, and generic result types use TypeToken/Class overloads that exist. The standalone (non-reactor) pom.xml is intentional and documented, and the lack of unit tests is consistent with the existing conformance-tests/ module and AGENTS.md/CONTRIBUTING.md (conformance handlers are deploy-only).

One confirmed, CI-breaking issue: the reusable-workflow reference still contains the literal placeholder @<ORCHESTRATOR_SHA> (see .github/workflows/otel-conformance-tests.yml:64). Because the workflow triggers on pull_request/push to sdk/**, otel-plugin/**, conformance-tests-otel/**, and pom.xml, merging as-is causes every matching PR to fail this check with an "invalid reusable workflow reference" error. It is acknowledged by the adjacent TODO, so the concrete gate is to pin a real SHA before this leaves draft/merges.

Minor (not blocking): OtelConformanceHandler's class Javadoc states the otel-execution suite runs "against the X-Ray backend only," but template.yaml deploys the execution functions in the same per-backend stacks as the invocation suite (Dash0/Datadog/S3 collector included), so that comment appears stale/inaccurate. Worth correcting to avoid misleading maintainers, but it has no runtime effect.

Residual test risk: none of the new handlers are exercised by local/replay tests (by design); their behavior is only validated by the external orchestrator once the SHA is pinned, so regressions in the ported scenarios would surface only in the deployed conformance run.

Reviewed commit 92b7102410345a19044b74782c7952056d32541b. Workflow run

Codex AI review

No actionable findings. Residual risk remains in the live SAM deployment and external OTel backend matrix, which were not executed under the read-only review constraints.

Reviewed commit f28b9a3401310d51b5b7f67e4d339d02abb5100e. Workflow run