[SDK] Add deepseek-v4-flash to VERIFIED_DEEPSEEK_MODELS (part of #3267)

[juanmichelini]

Parent issue

Sub-issue of #3267 — split out per @juanmichelini's request on the parent issue.

Scope

Issue 1 from the parent report: the model list / verified-models registry is missing the latest DeepSeek V4 models.

Current state (from triage on main, post-v1.23.1)

The dropdown UI itself lives in the OpenHands main repo, not in this SDK. What this repo exposes is VERIFIED_MODELS, served via the agent-server /llm/models/verified endpoint.

• deepseek-v4-pro was added to VERIFIED_DEEPSEEK_MODELS in Add DeepSeek-V4-Pro to verified models #3314 ✅
• deepseek-v4-flash is still missing from verified_models.py — looks like an oversight (it was added to model_features.py in Add DeepSeek-V4-Flash model #2979 but never to the verified list).

Whether the GUI dropdown actually consumes this endpoint vs. a hardcoded frontend list is an OpenHands-repo question and is tracked separately.

Proposed SDK-side fix

Add "deepseek-v4-flash" to VERIFIED_DEEPSEEK_MODELS in:

openhands-sdk/openhands/sdk/llm/utils/verified_models.py

A one-line addition should close the SDK-side half of this report.

Acceptance criteria

• deepseek-v4-flash appears in VERIFIED_DEEPSEEK_MODELS.
• /llm/models/verified returns it from the agent-server.
• Existing tests still pass; a small assertion in the verified-models tests covers the addition.

---

This sub-issue was filed by an AI agent (OpenHands) on behalf of @juanmichelini, following his instructions on the parent issue.