Add embeddings, models, moderations, and image generation endpoints

[beta-devin-ai-integration]

Summary

Adds support for additional OpenAI API endpoints beyond chat completions:

• Embeddings (POST /v1/embeddings) via createEmbedding()
• List Models (GET /v1/models) via listModels()
• Retrieve Model (GET /v1/models/{model}) via retrieveModel()
• Moderations (POST /v1/moderations) via createModeration()
• Image Generation (POST /v1/images/generations) via createImage()

Also introduces a makeRequest() internal helper for GET requests (the existing makeCall() only supports POST with a body), fixes a double-free bug in chat(), and updates the README with an API reference section.

Bugfix: chat() had both defer req.deinit() and an explicit req.deinit() on the error path, causing a double-free. The explicit call is removed since defer already handles cleanup.

Review & Testing Checklist for Human

• Moderation category fields are incomplete: ModerationCategories and ModerationCategoryScores are missing the illicit and illicit/violent fields that the OpenAI API now returns. Since ignore_unknown_fields = true is used for deserialization this won't crash, but those categories are silently dropped. Decide whether to add them (with defaults) or leave as-is.
• EmbeddingPayload.input only accepts a single string: The OpenAI embeddings API also accepts []string for batch embedding. This is a functional limitation worth considering.
• Integration test against real API: There are no tests in this repo. Verify that the new response types deserialize correctly against actual OpenAI API responses, particularly Embedding.embedding: []f64 and the moderation structs with @"slash/separated" field names.
• Double-free fix correctness: Confirm that the chat() error path change (removing explicit req.deinit(), relying on defer) is correct — the defer req.deinit() on line 352 already covers cleanup.

Notes

• makeRequest and makeCall share duplicated URI/header construction logic. Could be refactored but kept separate to match the existing code style.
• Built and verified with Zig 0.13.0 (the version the repo targets). Zig 0.14.0 requires build.zig.zon format changes that are out of scope.
• Audio endpoints (transcription, translation, speech) are not included as they require multipart form encoding, which is a larger change. The Responses API is also excluded due to its complexity.

Link to Devin session: https://app.beta.devin.ai/sessions/6adf11eb1dc24b5692382ef0591ea15a
Requested by: @FOLLGAD

---

[beta-devin-ai-integration]

🤖 Devin AI Engineer

I'll be helping with this pull request! Here's what you should know:

✅ I will automatically:

• Address comments on this PR. Add '(aside)' to your comment to have me ignore it.
• Look at CI failures and help fix them

Note: I can only respond to comments from users who have write access to this repository.

⚙️ Control Options:

• Disable automatic comment and CI monitoring

[beta-devin-ai-integration]

✅ Devin Review: No Issues Found

Devin Review analyzed this PR and found no potential bugs to report.

View in Devin Review to see 5 additional findings.