docs(datasets): document on_schema_change schema-evolution policy

[claudespice]

Summary

spiceai/spiceai#11261 wires up the dataset-level on_schema_change policy (the config was added in #10908 but previously unwired). Accelerated datasets can now evolve their schema in place for lossless, widening-compatible source changes instead of failing the refresh. This adds a reference section for the field in the dataset spec — it was undocumented (only mentioned in release notes).

Documents the four policy values verified against source:

• block (default) — OnSchemaChange::Block, preserves historical behavior
• fail
• append_new_columns
• sync_all_columns

Verified against crates/spicepod/src/component/dataset.rs (top-level dataset field, #[serde(rename_all = "snake_case")], default Block) and crates/runtime/src/init/dataset.rs (accelerated-datasets-only; non-accelerated datasets warn and treat it as inert). Engine support (DuckDB/SQLite/Turso/Cayenne in place; others degrade on restart) and the mode: file_update escape hatch are documented per the PR's implementation notes.

Addition → vNext only (website/docs/); older versions don't have the wired feature.

Source PRs

• feat(acceleration): widening-only schema evolution via on_schema_change spiceai#11261 — feat(acceleration): widening-only schema evolution via on_schema_change

Test plan

• cd website && npm run build passes
• Versioned-docs propagation: addition, vNext only — no versioned copies
• Files updated: 1 (website/docs/reference/spicepod/datasets.md)

[github-actions]

✅ Pull with Spice Passed

Passing checks:

• ✅ Title meets minimum length requirement (10 characters)
• ✅ Has at least one of the required labels: area/blog, area/docs, area/cookbook, dependencies
• ✅ No banned labels detected
• ✅ Has at least one assignee: claudespice

[github-actions]

🚀 deployed to https://bbba7a95.spiceai-org-website.pages.dev