fix(openaimodel): send executable code parts as text - #1956

Please ensure you have read the contribution guide before creating a pull request.

Linked issue

• Closes: model/openaimodel: send executable code and code execution results as text #1924

Problem:
A Gemini turn that stores ExecutableCode or CodeExecutionResult makes the next OpenAI request fail before it is sent. Both Chat Completions and Responses reject the part with openai: unsupported content part: ExecutableCode (or CodeExecutionResult).

Solution:
Send those parts as text, using the same fences as adk-python's Responses converter: a python block for the code and a code_output block for the result. Both APIs do this. adk-python's Chat Completions converter drops these parts; Go deliberately preserves them there so a cross-model session does not hide the code or its output from the model.

Behavior change

A request that used to fail on these parts now includes them as text. Other unsupported parts are still rejected.

Testing Plan

Unit Tests:

• All unit tests pass locally.
• Both endpoints assert the exact request payload for code-only and code-plus-image histories, including the media-message path.

go test -count=1 -race -timeout 180s ./model/openaimodel/...

ok  google.golang.org/adk/v2/model/openaimodel
ok  google.golang.org/adk/v2/model/openaimodel/internal/completions
ok  google.golang.org/adk/v2/model/openaimodel/internal/responses
ok  google.golang.org/adk/v2/model/openaimodel/internal/shared

golangci-lint run (v2.3.1) on those three internal packages reported 0 issues.

With your source change reverted and your tests kept, which test fails?

Leaving the new tests in place and not passing PartFieldCodeExecution from the converters:

• TestBuildParams_CodeExecutionAsText (responses and completions)
• TestBuildParams_DropsReplayedThoughts/thought_text_riding_on_code_drops_only_the_text

They fail with openai: unsupported content part: ExecutableCode.

Removing the media-path insertion instead makes these focused cases fail by omitting the fenced code before the image:

• TestBuildParams_MediaOnTheWire/executable_code_before_an_inline_image_keeps_its_order (Chat Completions)
• TestBuildParams_CodeExecutionAsText/code_keeps_its_place_on_a_media_message (Responses)

Manual End-to-End (E2E) Tests:

No live model call. The failure is in request construction, and the tests above check the payload that would be sent.

Checklist

• I have read the CONTRIBUTING.md document.
• I have performed a self-review of my own code.
• I have commented my code, particularly in hard-to-understand areas.
• I have added tests that prove my fix is effective or that my feature works.
• New and existing unit tests pass locally with my changes.
• I have manually tested my changes end-to-end.
• Any dependent changes have been merged and published in downstream modules.