Add Perspective.js artifact integration

[hachej]

Summary

• add provider-neutral Perspective.js dataset/artifact helpers for BSL query results
• add SemanticTable convenience methods for Perspective dataset and artifact export
• separate inline datasets (kind: bsl.perspective.dataset, data) from static artifact manifests (kind: bsl.perspective.artifact, data_ref)
• add safe artifact-id validation, JSON-safe payload/viewer sanitization, Decimal-aware schema inference, and semantic role preservation
• document the Perspective integration and add optional viz-perspective extra
• add focused tests plus scripts/test_perspective_local.py smoke script

Tests

• uv run ruff check src/boring_semantic_layer/integrations src/boring_semantic_layer/tests/test_perspective_integration.py src/boring_semantic_layer/expr.py scripts/test_perspective_local.py
• uv run pytest src/boring_semantic_layer/tests/test_perspective_integration.py
• uv run python scripts/test_perspective_local.py

Reviews

• normal reviewer pass: no blockers/actionable findings
• thermo-nuclear code quality review: no maintainability blockers; ready to merge

Note: broader chart test run was attempted earlier, but optional chart dependencies (altair / plotly / plotext) are not installed in this environment, so those unrelated tests fail with ModuleNotFoundError.

[hachej]

Claude Code Opus/ultracode review iteration:

• Initial review found Arrow/manifest schema divergence for Decimal values and nullable integer/date schema fidelity gaps.
• Fixed by coercing JSON and Arrow outputs from the inferred Perspective schema, normalizing Decimal to float consistently and preserving nullable integer/date semantics.
• Added inline data_format to dataset payloads and updated Arrow install guidance to mention the viz-perspective extra.
• Final Claude Code review verdict: clean, no remaining blocker/high/medium actionable issues.

Validation after fixes:

• uv run pytest src/boring_semantic_layer/tests/test_perspective_integration.py
• uv run python scripts/test_perspective_local.py