dsh-speech-input: Add a Microphone Button to the DeepSeek Harness Web Input Field

# Preface In dsh web, conversing with a model often requires typing out long prompts for several minutes. Since the input box lacks a voice input entry point, dictation usually relies on system dictation and pasting, or directly modifying the Harness DOM to inject controls—both of which are unreliable, with the latter specifically being prone to breaking after page updates. The `dsh-speech-input` introduced below takes a different approach: it adds a native microphone button to the right of the input box using official slots. Clicking it starts dictation, writing recognition results to the current draft in real-time. Clicking it again stops the process and appends punctuation to the end of the sentence, ensuring the message is never sent automatically. DSH's philosophy is "Everything is a plugin," and input box enhancement falls squarely within the scope of the plugin system. # What is this `dsh-speech-input` is lizne