Add v2 page-memory schema support

[suguanYang]

Summary

• add SDK-level apiVersion support with default v1 behavior and per-call overrides for jobs, documents, polling/loading, and retrieval
• add v2 page-memory schemas: page chunks, contentSource, page metadata/entities, pageChunks, and retrieval chunkTypes / expanded dataType
• update result parsing, local knowledge helpers, tests, README examples, and add a patch changeset for the SDK release

Notes

• This is intended to publish as a patch release after merge so downstream ^0.10.0 consumers can pick up v2/page-memory support.
• The API-side schema work is in Add v2 page-memory ingestion API knowhere#196.

Verification

• pnpm run format:check:all
• pnpm run lint:all (passed with existing warning-only lint output)
• pnpm run typecheck:all
• pnpm run test:all (278 root tests + 11 MCP tests)
• pnpm run build:all
• git diff --check

[suguanYang]

Follow-up update for the v2-only direction:

• Removed the public apiVersion / v1 parse-flow switching surface from the SDK.
• SDK requests now always use /v2 for parse, jobs, documents, polling, and retrieval.
• @ontos-ai/knowhere-sdk and @ontos-ai/knowhere-mcp are now explicitly versioned as 2.0.0; I removed the normal Changesets major entry because a major bump from 0.x would have released 1.0.0, not 2.0.0.
• Legacy parsed-result chunk inspection remains for existing v1 result ZIPs.

Validation:

• pnpm run format:all
• pnpm run format:check:all
• pnpm run lint:all (existing warnings only)
• pnpm run typecheck:all
• pnpm run test:all
• pnpm run build:all
• pnpm run release:check-changeset -- origin/main
• git diff --check

Remote CI is green for Node 20 and Node 22 on commit 31c88e2.