v1.0.2: Critical Stability, Security & Model Pipeline Fixes

Release Highlights in v1.0.2

• Incremental Audio Streaming: Implemented incremental sample consumption in AudioService and continuous stream feeding to Vosk ASR, eliminating the ~33x audio duplication issue.
• Hardware & Model Availability Gatekeeper: Enhanced ChallengeRegistry to verify offline ML models exist on disk before selecting ML challenges, falling back cleanly to timing.reaction.
• Unified Asset & Model Resolution: Added sudo_fun.core.assets with prioritized discovery across environment variables, user share, and repo directories.
• Environment Variable Isolation: Added restore_pristine_environment() to prevent Wayland and Qt environment mutations from leaking into privileged commands executed under sudo.
• Portability: Added dynamic sudo binary resolution in executor.py supporting NixOS setuid wrappers (/run/wrappers/bin/sudo), PATH, /usr/bin/sudo, and /bin/sudo.
• Security: Switched all asset downloads in install.sh and release.yml to HTTPS.
• Installer Enhancements: install.sh now automatically downloads and verifies offline ML models.
• Expanded Test Suite: Added unit tests for assets, audio streaming, environment hygiene, and model fallback (29 tests passing).