feat(ci): add Bedrock integration tests with record/replay

[skamenan7]

What does this PR do?

Adds Bedrock integration tests to CI using a record/replay mechanism. Tests run against pre-recorded API responses, without the need for AWS credentials in CI.

The main challenge was that Bedrock's OpenAI-compatible API doesn't support everything - no tool calling, no embeddings, no dynamic model listing. So instead of running the full base suite (which would fail on ~40 tests), I created a dedicated bedrock suite with just the tests that actually work.

Changes:

• New run-bedrock.yaml stack config with dummy API key for replay mode
• bedrock suite in suites.py pointing to 3 test functions (6 parametrized tests total)
• Config resolution fix for distro::file.yaml format in library mode
• Added a docs/source/providers/inference/bedrock_recording_guide.md so contributors with AWS access can re-record tests when needed

Closes #4095

Test Plan

Run from tests/integration/inference:

  uv run pytest -v \
    test_openai_completion.py::test_openai_chat_completion_non_streaming \
    test_openai_completion.py::test_openai_chat_completion_streaming \
    test_openai_completion.py::test_inference_store \
    --setup=bedrock \
    --stack-config=ci-tests::run-bedrock.yaml \
    --inference-mode=replay \
    -k "client_with_models"

Expected: 6 passed

test_openai_completion.py::test_openai_chat_completion_non_streaming[client_with_models-txt=bedrock/openai.gpt-oss-20b-1:0-inference:chat_com
 pletion:non_streaming_01] PASSED [ 16%]
 test_openai_completion.py::test_openai_chat_completion_streaming[client_with_models-txt=bedrock/openai.gpt-oss-20b-1:0-inference:chat_complet
 ion:streaming_01] PASSED [ 33%]
 test_openai_completion.py::test_inference_store[client_with_models-txt=bedrock/openai.gpt-oss-20b-1:0-True] PASSED                          [
  50%]
 test_openai_completion.py::test_openai_chat_completion_non_streaming[client_with_models-txt=bedrock/openai.gpt-oss-20b-1:0-inference:chat_com
 pletion:non_streaming_02] PASSED [ 66%]
 test_openai_completion.py::test_openai_chat_completion_streaming[client_with_models-txt=bedrock/openai.gpt-oss-20b-1:0-inference:chat_complet
 ion:streaming_02] PASSED [ 83%]
 test_openai_completion.py::test_inference_store[client_with_models-txt=bedrock/openai.gpt-oss-20b-1:0-False] PASSED
 [100%]

[derekhiggins]

remote::bedrock is part of ci-tests/run.yaml , can this be used instead (i.e. no need for a new run.yaml)

[skamenan7]

Good point! I'll check if ci-tests/run.yaml can be used directly. The separate file was created to have a minimal bedrock-only config, but if the existing run.yaml already includes bedrock provider, I can simplify by using that instead.

[derekhiggins]

I'm not sure we can just add this to CI here, anybody updating tests that need new recording would need a bedrock key (or we'd need a key in CI for new recordings, is that the plan?)