Add 10 preloaded datasets on first app start

[Irozuku]

Summary

Adds automatic dataset seeding on first app start. On the first run, DashAI copies 10 pre-processed datasets (Arrow format) from a bundled zip into the local datasets directory and registers them in the database (no user action required). Subsequent starts skip seeding via a .seeded sentinel file.

---

Type of Change

Check all that apply like this [x]:

• Backend change
• Frontend change
• CI / Workflow change
• Build / Packaging change
• Bug fix
• Documentation

---

Changes (by file)

• DashAI/back/seeds/__init__.py: seeding logic - checks sentinel, extracts zip, copies dataset folders, inserts DB rows with FINISHED status
• DashAI/back/seeds/manifest.json: maps dataset names to row/column counts used for DB registration
• DashAI/back/seeds/seed_datasets.zip: 10 pre-processed datasets in Arrow format (ai-vs-human, languages, urdu-depression-data, students, spanish-mnist, food-menu, energy, possum, auditory-skills, cifar10-subset)
• DashAI/back/app.py: calls seed_datasets_if_first_run() after DB migrations on startup
• dashai.spec: includes DashAI/back/seeds in PyInstaller datas so the zip ships with the executable
• MANIFEST.in: includes DashAI/back/seeds so the zip ships with the pip package

---

Testing

• Install via pip install on Windows, macOS and Linux, verify 10 datasets appear on first start
• Run executable (PyInstaller build) on Windows, verify 10 datasets appear on first start
• Run executable (PyInstaller build) on macOS, verify 10 datasets appear on first start
• Restart app after first run, verify seeding is skipped (no duplicates)
• Upload a dataset with the same name as a seed dataset before first run, verify no conflict

---

Notes

• Seeding only runs once - sentinel file ~/.DashAI/.seeded is written after completion
• If zip is missing (e.g. dev environment without assets), startup continues normally with a warning
• Datasets with the same name already in the DB are skipped without touching the filesystem