feat: Add single-result modes to Assets queries

[kof]

Summary

• Add many, one, first, and last result modes to Assets queries while keeping many as the backward-compatible default.
• Return direct items for single-result modes, validate deterministic ordering, and report duplicate one matches as a structured MULTIPLE_RESULTS error.
• Update Builder controls, preview/published runtime handling, SDK resource encoding, MCP/CLI guidance, generated contracts, and fixtures.

Implementation TODOs

• Add result-mode schemas and core execution behavior.
• Select single results before content hydration and preserve deterministic ID tie-breaking.
• Add structured duplicate-match transport errors.
• Update Builder controls and hide/reset pagination for single-result modes.
• Keep existing settings first, then Output and Result, without a duplicate Return label.
• Preserve legacy saved resources with an implicit many default.
• Update preview, published runtime, SDK, protocol, MCP, and CLI behavior.
• Regenerate runtime contracts, server metadata, CLI docs, and fixtures.
• Review documentation impact and update the authoritative MCP/CLI guidance.
• Add regression coverage and verify the new tests fail without the implementation.
• Run the agent evaluation suite (not run; requires explicit user approval).

Compatibility and behavior

• Existing queries without result continue to behave as many and retain the ID-keyed collection shape.
• one returns an item or null, but fails when more than one asset matches.
• first and last require explicit sorting; the existing asset-ID tie-breaker makes equal sort values deterministic.
• Single-result modes ignore and do not persist pagination controls. Materialized collection queries remain collection-only; single-result queries execute dynamically.

Verification

• Content engine typecheck and full test suite (291 tests).
• Query builder, query-builder React, SDK, protocol, asset uploader, project build, Builder, and CLI targeted typechecks/tests.
• Root lint.
• MCP documentation tests, typecheck, sync check, and generated-doc checks.
• Generated API contract checks.
• Fixture generation pipeline.
• Final focused content-engine regression run (53 tests).
• Final DRY review: affected typechecks, focused tests, lint, package boundaries, generated checks, and fixtures.

Closes #6039