test(otel): observe the SDK's default OTLP protocol instead of hardcoding it - #8256

Motivation

The OTLP protocol tests hardcoded each SDK's default protocol per signal in utils.py. When an SDK migrates its default (.NET is doing so), the tests have to be disabled, updated and re-enabled.

Changes

The default is now observed at runtime instead of hardcoded. How the pieces fit:

• default_protocol (fixture, otel_env_vars/conftest.py): returns a callable, default_protocol(signal). On first use it starts the SDK through test_library_factory with no protocol variable set and exports once to the HTTP listener (the content type gives http/protobuf or http/json), then falls back to the gRPC listener (grpc). The result is cached per signal and worker. Because it is a callable, it only runs when a test needs the default, so explicit-protocol tests don't depend on it.
• sdk_default / sdk_non_default (utils.py): functions of (signal, default) used as parametrize values for inputs that depend on the default. sdk_non_default picks a transport the SDK supports that isn't its default; the case-insensitive tests call .upper() on it.
• resolve_protocol (utils.py): turns a parametrize value into the env value. Functions are called with the observed default; literals pass through.
• expected_protocol (utils.py): resolves protocol and generic_protocol, then applies precedence: signal variable, generic variable, then the SDK default (unset, empty or invalid).
• expected_protocol (fixture, each test file): calls the util with the test's parametrize values. library_env uses it to point the endpoint at the HTTP or gRPC listener, and the test uses it for the content-type assertion, so both always agree.

Traces get the same treatment, with DD_TRACE_OTEL_ENABLED, DD_TRACE_OTEL_SEMANTICS_ENABLED and OTEL_TRACES_EXPORTER=otlp. test_unset_uses_specification_default still asserts http/protobuf. Test IDs are unchanged, so no manifest updates.

Empty and invalid values are now compared against the SDK's actual default instead of a documented one, so these XPASS:

• Go logs test_empty / test_invalid (APMAPI-2501): Go uses http/protobuf while documenting http/json, which test_documented_json_default still catches.
• .NET traces test_empty_is_treated_as_unset (APMAPI-2531): .NET defaults to http/json, which test_unset_uses_specification_default still catches.

Verified locally on dotnet 3.57.0, and on golang 2.13.0-dev and php 1.24.0 before the final helper refactor.

Workflow

1. ⚠️ Create your PR as draft ⚠️
2. Work on you PR until the CI passes
3. Mark it as ready for review
   • Tests, manifest, weblog are modified -> you'll need a review from system-tests-reviewers: ask to one of youre co-worker familiar with the tested feature.
   • Framework is modified, or non obvious usage of it -> get a review from system-tests-core (slack)

🚀 Once your PR is reviewed and the CI green, you can merge it!

🛟 #apm-shared-testing 🛟

🤖 Generated with Claude Code