docs: Replace load: on_demand with deferred dataset initialization

[claudespice]

Summary

• Replace the removed load: on_demand section with documentation for the new deferred dataset initialization feature
• Update the ready_state section to cross-reference deferred initialization when combined with columns[].type

Source PRs

• feat(runtime): declared-schema deferred datasets spiceai#10669 — feat(runtime): declared-schema deferred datasets

Changes

• website/docs/reference/spicepod/datasets.md: Replaced the ## load section (documenting the removed load: on_demand option) with a ## Deferred dataset initialization section that documents the new approach using ready_state: on_registration + fully declared columns[].type. Added a breaking change warning. Updated the ready_state section to mention deferred initialization.

Test plan

• Verified the breaking change warning is prominent
• Links are valid (cross-references within the same page)
• Only vNext docs updated (feature is new, not present in versioned docs)

[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://485edf19.spiceai-org-website.pages.dev