fix speech auth

[lzc850612]

Purpose

What is the context of this pull request? Why is it being done?

close #2073

Changes

Are there any changes that need to be called out as significant or particularly difficult to grasp? (Include illustrative screenshots for context if applicable.)

In speech scenario, when Auth fails the dialog doesn't end properly. This PR is to fix that.

Tests

Is this covered by existing tests or new ones? If no, why not?

Feature Plan

Are there any remaining steps or dependencies before this issue can be fully resolved? If so, describe and link to any relevant pull requests or issues.

Checklist

General

• I have commented my code, particularly in hard-to-understand areas
• I have added or updated the appropriate tests
• I have updated related documentation

Bots

• I have validated that new and updated responses use appropriate Speak and InputHint properties to ensure a high-quality speech-first experience
• I have replicated language model changes across the English, French, Italian, German, Spanish, and Chinese .lu files and validated that deployment is successful

Deployment Scripts

• I have replicated my changes in the Virtual Assistant Template and Sample projects
• I have replicated my changes in the Skill Template and Sample projects