[docs] Reorganize append table guides and replace diagrams with SVGs

[JingsongLi]

Purpose

Append-table guidance is spread across overlapping pages, and the overview mixes setup, streaming, query optimization, and row-level behavior. Reorganize it into seven focused guides with a clear sidebar and consistent Flink/Spark examples, while preserving existing page URLs and overview anchors.

• Explain layout selection, compaction and incremental clustering, query pruning and indexes, row-level operations, and row tracking with their configuration requirements and behavioral limits.
• Add partition-scoped clustering examples and system-table queries for checking clustering results and index coverage.
• Replace the two PNG diagrams with six accessible SVG diagrams covering streaming compaction, bucket ordering, clustering, file pruning, row tracking, and deletion vectors.
• Align the Flink overwrite-reading guidance with append-table behavior.

Tests

• Docusaurus production build passed (node node_modules/@docusaurus/core/bin/docusaurus.mjs build from docs).
• REST OpenAPI validation passed: 56 catalog operations and 6 management operations.
• Checked 113 internal links/anchors, all six SVG references and XML accessibility metadata, and preserved overview anchors across the seven generated pages.
• Inspected rendered pages and diagrams in the browser and verified synchronized engine tabs.
• git diff --check passed. SQL examples were checked against implementation and existing tests; no Spark or Flink jobs were executed.

[leaves12138]

Reviewed at 0031d82. One documentation correction is needed before approval: the partition-scoped clustering examples need to explain how historical auto-clustering expands the work beyond the requested partitions. See the inline comment.

Independent validation passed:

• Production documentation build and REST OpenAPI validation.
• All 82 unique internal body links/anchors in the eight changed generated documentation pages.
• All five existing overview section anchors are preserved; the sidebar retains the original guides and adds the three new guides.
• Browser checks of all seven append-table pages, including images, engine tabs, 390 px mobile width, and JavaScript page errors.
• All six SVGs pass XML/accessibility-metadata and rendered-text-bound checks; the rendered diagrams were visually inspected. No references to the two removed PNGs remain in the documentation tree.

Streaming overwrite behavior, bounded-watermark handling, row tracking, row-level operations, and clustering configuration were checked against the implementation and existing test assertions. SQL examples and Java/engine tests were not executed.

[leaves12138]

LGTM. Re-reviewed at 20edd15.

The previous finding is fully addressed. Both Spark and Flink examples now distinguish the requested partition scope from additional historical clustering. The historical-clustering section correctly states the explicit-predicate and idle-duration prerequisites, the additional-partition limit, the full-clustering behavior for those extra partitions even with a minor foreground strategy, and how to keep the job strictly scoped. These statements match the implementation and existing test assertions.

Only this guide changed since my previous review; no additional findings.

Revalidation on the updated head:

• Production documentation build and REST OpenAPI validation pass with freshly installed dependencies.
• All 82 unique internal body links/anchors across the eight changed pages resolve, and all five original overview section anchors remain available.
• Sidebar and SVG metadata checks pass.
• Browser checks confirm both engine-specific scope warnings, navigation to the historical-clustering section, image loading, and 390 px mobile layout, with no JavaScript page errors.

Both GitHub checks are successful. SQL examples and Java/engine tests were not executed; this remains a documentation-only change.