Add --download-models flag to run.sh for automatic model download

[Copilot]

README claimed run.sh is an "all-in-one" setup method but required separate manual model downloads. Models are not included in the repository (only .gitignore files in assets/).

Changes

Created download_models.py

• Standalone script to download all required models without starting web UI
• Checks existing models first, skips if present
• Exit codes: 0=success, 1=error, 2=warning (partial download)
• Uses existing infer.lib.rvcmd infrastructure

Enhanced run.sh

• Added --download-models flag to trigger model download before startup
• Proper exit code handling (exits on error, continues on warning)
• User confirmation before starting UI after download

Updated README

• Documented ./run.sh --download-models for first-time setup
• Added standalone python download_models.py option
• Clarified "Assets" section with automatic vs manual download methods
• Removed misleading claim about included models

Usage

# First time
./run.sh --download-models

# Subsequent runs
./run.sh

# Standalone download
python download_models.py

Model download uses the existing RVC-Models-Downloader integration already present in infer/lib/rvcmd.py.

---

💬 We'd love your input! Share your thoughts on Copilot coding agent in our 2 minute survey.