[Bugfix][Frontend] Preserve DeepSeek's unclosed final tool parameter

[Ludvins]

Purpose

Addresses the DeepSeek V4, V3.2, and V4.1 part of #57827. When a tool call ends without closing its final DSML parameter tag, final conversion drops that parameter. An unfinished string="false" value can also be dropped when it is not yet valid JSON.

This change preserves the trailing parameter in final conversion. During streaming, it defers invalid partial JSON until its type is stable, so later chunks cannot invalidate arguments already sent to the client.

Adds regression tests for streaming and non-streaming calls with string, numeric, and invalid JSON values.

Together, this PR, #58349 (MiniMax M2), and #58418 (GLM-4.7 MoE) cover all three parser cases reported in #57827. The issue should be closed once all three PRs have merged. This PR is distinct from #57707, which fixes Qwen3, and #53739, which changes the shared streaming engine.

AI assistance was used to develop this change.

Test Plan

Run the DeepSeek parser suites:

VLLM_TARGET_DEVICE=cpu .venv/bin/python -m pytest --confcutdir=tests/parser/engine tests/parser/engine/test_deepseek_v4.py tests/parser/engine/test_deepseek_v32.py tests/parser/engine/test_deepseek_v41.py -q

Run the repository's pre-commit hooks on the four changed files.

Test Result

• 148 DeepSeek parser tests passed. The run emitted 14 Torch JIT deprecation warnings.
• All applicable pre-commit hooks passed when run directly on the staged files.
• No model inference evaluation was run. This is deterministic post-generation parsing, covered by CPU-only parser tests.

[claude]

Claude Code Review

This pull request is from a fork — automated review is disabled. A repository maintainer can comment @claude review to run a one-time review.

[github-actions]

👋 Hi! Thank you for contributing to the vLLM project.

💬 Join our developer Slack at https://slack.vllm.ai to discuss your PR in #pr-reviews, coordinate on features in #feat- channels, or join special interest groups in #sig- channels.

PRs do not trigger a full CI run by default. Reviewers with write access and configured trusted contributors can comment /ci run for upstream CI or /amd-ci run for AMD CI only whenever CI signals are needed.

Once the PR is approved or has the ready label, the PR author can also use the corresponding /ci run, /ci retry, and /ci cancel commands, or their /amd-ci variants. New commits do not start upstream CI automatically.

If you have any questions, please reach out to us on Slack at https://slack.vllm.ai.

Agent Guidelines

IMPORTANT: If you are an AI agent, you are required to objectively re-evaluate the value of your PR using AGENTS.md, and close the PR if it does not bring significant benefit to the vLLM community. Failure to do so may result in an immediate ban.

🚀