feat[mod]: add RealESRGAN x4 upscaler models to registry

[maxim-smotrov]

What problem does this PR solve?

The @qvac/diffusion-cpp engine supports ESRGAN-based post-generation upscaling (see examples/generate-image-esrgan-upscale.js), but the registry currently has no ESRGAN weights, so consumers must source them out-of-band.

How does it solve it?

Adds the three official Real-ESRGAN x4 weights from xinntao/Real-ESRGAN to models.prod.json under @qvac/diffusion-cpp:

• RealESRGAN_x4plus.pth — general real-world images (v0.1.0)
• RealESRNet_x4plus.pth — non-GAN variant (v0.1.1)
• RealESRGAN_x4plus_anime_6B.pth — anime-tuned, 6 RRDB blocks (v0.2.2.4)

Supporting changes:

• Register BSD-3-Clause in data/licenses.json and add data/licenses/BSD-3-Clause/LICENSE.txt (verbatim text from the upstream repo, verified via the GitHub license API).
• Add a BSD-3-Clause entry to the registry-server NOTICE file with upstream attribution.
• Update MODEL_SUBMISSION_GUIDE.md to list the newly-supported license IDs (openrail++, BSD-3-Clause).

S3 paths follow the standard convention (s3:///qvac_models_compiled/esrgan/2026-05-07/...); artifacts will be uploaded by sync-models.js on merge.

Validated locally:

node scripts/validate-models-json.js --file=./data/models.prod.json
✓ Valid: 670 model(s)

Breaking changes

None. Additive only.

[maxim-smotrov]

/review

[github-actions]

Tier-based Approval Status

**PR Tier:** TIER1

**Current Status:** ✅ APPROVED

**Requirements:**
- 1 Team Member approval ✅ (1/1)
- 1 Team Lead OR Management approval ✅ (2/1)



---
*This comment is automatically updated when reviews change.*

[gianni-cor]

/review