fix(Qwen3/GLM5.2): exclude frontend-undecodable padded vocab rows

[Kokoro2336]

Description

Fixes #680

Type of Change

• Bug fix (non-breaking change which fixes an issue)

Checklist

• My code follows the style guidelines of this project (see docs/conventions/coding-style.md).

• I have performed a self-review of my own code.

• I have formatted my commits according to Commitizen conventions.

• I have run the local test suite and all tests pass (see CLAUDE.md).

[chatgpt-codex-connector]

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: bc62f3e093

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

• Open a pull request for review
• Mark a draft as ready
• Comment "@codex review".

If Codex has suggestions, it will comment; otherwise it will react with 👍.

Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".

[chatgpt-codex-connector]

Apply the selectable vocab to Qwen3 target logits

When a Qwen3 tokenizer exposes fewer decodable IDs than the checkpoint vocab_size (the case this branch detects and logs here), the normal target path still projects and samples the full checkpoint vocab: LocalQwen3Lane::new sizes decode logits and SampleScratch with config.vocab_size, select_step_tokens samples those logits, and DFlash verify uses the same target logits. As a result, non-speculative Qwen3 requests and the verify bonus path can still choose the padded, frontend-undecodable rows; this new selection_vocab value only bounds the draft/Markov side unless the main prefill/decode/verify logits use it too.

Useful? React with 👍 / 👎.