refactor: Fase 1 — Consolidação Técnica para Beta (v0.2.0)

[Copilot]

Features v0.2 were built on top of accumulated technical debt — 3 duplicate CLIs, pipeline logic copied 4–6×, phantom dependencies, dead modules, 2 broken scripts, and a GuruMatrix only 30% populated. This PR cleans the foundation before further development.

Pipeline & CLI

• Extracted src/gurudev/pipeline.py — single source of truth for compilar(), executar(), RECURSOS_DEMO
• cli.py now imports from pipeline.py; removed popular_minimo() call from executar() (redundant with _carregar_dados_v02())
• Deleted src/cli/gurudev_cli.py (argparse legacy), gurudev-cli.py (root copy), and src/cli/ directory
• Removed gurudev-cli entry point from pyproject.toml

Dead Code & Phantom Dependencies

• Removed pydantic>=2.0 (zero imports across the codebase)
• Deleted src/gurudev/logger.py (singleton logger, never imported anywhere)
• Deleted ResourceNotFoundError from exceptions.py and parse_tipo() from parser.py
• Deleted scripts/generate_embeddings.py (imports openai, not a dep) and scripts/populate_priority_domains.py (imports nonexistent gurumatrix.expansion_tool)

GuruMatrix: 30 → 60 cells

Added 3 new domains (CIENCIA, MATEMATICA, LINGUAGEM) × 10 ontologies each, with semantic relations where meaningful:

"ACAO_CIENCIA": {
  "objetos": ["fatorial", "fft", "média", "derivada", "integral"],
  "relacoes": ["equivalencia", "similitude"],
  "instrucoes": ["EVALUATE", "APPLY", "COMPARE"]
}

Updated embeddings_v0.2.json with deterministic 10-float vectors for all 30 new cells.

Cleanup

• Removed popular_minimo() from GuruMatrix (data already in data_v0.2.json); tests updated accordingly
• Removed [[tool.mypy.overrides]] with ignore_errors = true (no longer needed — 0 mypy errors)
• Fixed tautological assertion in test_density.py (else True branch made the check a no-op)
• Bumped version 0.1.0 → 0.2.0