Reusing Latent Speech Representations for Query-Conditioned Topic Localization in Transcripts | TickerVault