test(bdd): add C++ SDK BDD tests for basic messaging

[seokjin0414]

What

Adds BDD coverage for the C++ SDK by implementing the shared
bdd/scenarios/basic_messaging.feature with cucumber-cpp and wiring it into the
Docker-based BDD harness and CI, the same way the other SDKs run. This is the
second part of #2965; the messaging FFI it relies on landed in #3046.

How it works

cucumber-cpp only supports the Cucumber wire protocol, so the step definitions
(bdd/cpp/features/step_definitions) build with CMake into a wire server, and a
Ruby Cucumber runner (cucumber + cucumber-wire, pinned to the versions
cucumber-cpp itself tests against) drives the feature file over the wire against a
real iggy-server. The step logic and assertions are all C++ and exercise the
SDK's FFI (get_streams, create_stream, create_topic, send_messages,
poll_messages). The Iggy C++ SDK is built with cargo and linked into the wire
server as a staticlib.

Heads-up for reviewers

Because cucumber-cpp is wire-protocol-only, the C++ BDD image pulls in a Ruby
runtime (cucumber + cucumber-wire). Flagging this explicitly in case it
changes how you'd like the harness shaped.

Notes

Everything lives under bdd/cpp, next to the other SDK BDD suites; foreign/cpp
is unchanged apart from the reserve() fix to the existing test.

The TearDown fixture and the magic-number cleanup discussed on Discord are tracked
separately by @slbotbm.

Refs #2965

[slbotbm]

@seokjin0414 the test should be in he root bdd/ folder, not inside of foreign/cpp

[slbotbm]

ignore the above, i didn't read the description😅

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 74.09%. Comparing base (307fdb1) to head (61e1c76).
⚠️ Report is 7 commits behind head on master.

Additional details and impacted files

@@             Coverage Diff              @@
##             master    #3554      +/-   ##
============================================
+ Coverage     74.07%   74.09%   +0.02%     
  Complexity      937      937              
============================================
  Files          1249     1258       +9     
  Lines        128248   131452    +3204     
  Branches     104116   107363    +3247     
============================================
+ Hits          94994    97396    +2402     
- Misses        30219    30914     +695     
- Partials       3035     3142     +107     

Components	Coverage Δ
Rust Core	74.76% <ø> (+0.05%)	⬆️
Java SDK	62.44% <ø> (ø)
C# SDK	71.41% <ø> (-0.65%)	⬇️
Python SDK	88.88% <ø> (ø)
PHP SDK	84.29% <ø> (ø)
Node SDK	91.26% <ø> (-0.10%)	⬇️
Go SDK	40.14% <ø> (ø)
see 89 files with indirect coverage changes

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.
• 📦 JS Bundle Analysis: Save yourself from yourself by tracking and limiting bundle sizes in JS merges.

[seokjin0414]

@slbotbm
no worries! ended up moving it to bdd/cpp anyway. i kept it out of foreign/cpp by building the wire server with CMake (FetchContent cucumber-cpp + the cxx staticlib) instead of Bazel, so it's self-contained like the other bdd dirs and
foreign/cpp now only carries the reserve() fix.
can switch it to a Bazel target under foreign/cpp if you'd prefer.

[slbotbm]

@seokjin0414 I didn't understand why did you use cmake instead of bazel. I think it would be possible to use bazel in bdd/cpp. you could define iggy_cpp as a bazel dependency in bdd/cpp/MODULE.bazel, and then locally define the path to iggy_cpp as ../../foreign/cpp.

[slbotbm]

/author

[seokjin0414]

@slbotbm the cmake route was just to keep the test self-contained in bdd/cpp without touching foreign/cpp — i didn't realize local_path_override could pull the sdk in cleanly. it can, so i switched to bazel as you suggested: bdd/cpp is its own module now, cucumber-cpp from the registry + iggy_cpp via local_path_override(../../foreign/cpp). cmake is gone.

the only change to foreign/cpp is one line — making the iggy-cpp target public so the harness can depend on it. build + the full wire e2e (17 steps) pass.

[slbotbm]

Let's add a .clang-format as well that copies the one from foreign/cpp

You can use /ready when your PR is ready for review again

[slbotbm]

Let's also commit the ruby lockfile as well.

[seokjin0414]

@slbotbm
also addressed the two non-inline ones: added bdd/cpp/.clang-format (copied from foreign/cpp), and committed bdd/cpp/Gemfile.lock for the ruby side. rebased onto master — CI is green.

[seokjin0414]

/ready

[slbotbm]

Let's change this to a 3-stage build: the first stage uses gcr.io/bazel-public/bazel:9.1.1 and builds bdd_wire_server using bazel, the second stage uses ruby:3.3-slim-trixie and installs the rube deps using bundler, and the third step also uses ruby:3.3-slim-trixie, copying the ruby deps + bdd_wire_server from previous stages and runs the test. This reduces the final image from ~10 gb to 330 mb on my machine

[seokjin0414]

done. 3-stage now: bazel-public:9.1.1 builds the wire server, ruby:3.3-slim-trixie installs the gems with bundler, and the runtime stage copies just the binary + vendored gems. final image is ~220 MB (was ~6.5 GB). e2e still green, 17/17 steps.

[seokjin0414]

/ready