docs: group user and contributor guide nav into captioned sections

[andygrove]

Which issue does this PR close?

Part of #4421. Lands step 1 of the migration plan (captions/grouping; no page moves or content changes).

Rationale for this change

The user guide and contributor guide toctrees had grown into flat 14- and 23-item lists with no narrative grouping. The issue proposes a section-based layout that makes it obvious where Getting Started lives, what is reference vs. operational vs. architectural, and where integrations and project mechanics fit. Step 1 is the toctree restructure, kept small so reviewers see the new shape without page renames or content rewrites.

The pydata-sphinx-theme used here builds the sidebar from the global TOC, so nested toctree captions get flattened away. Without a sidebar fix the new captions wouldn't render anywhere visible, so this PR also adjusts docs-sidebar.html to render the per-section captions in the left nav.

What changes are included in this PR?

docs/source/user-guide/latest/index.rst — split the single 14-item toctree into 5 captioned toctrees:

• Getting Started — Installing Comet, Configuration Settings
• What Comet Supports — Supported Data Sources, Data Types, Operators, Expressions, ScalaUDF and Java UDF Support, Compatibility Guide
• Operating Comet — Understanding Comet Plans, Tuning Guide, Metrics Guide
• Integrations — Iceberg Guide, Kubernetes Guide
• Advanced — Building From Source

docs/source/contributor-guide/index.md — split the single 23-item toctree into 7 captioned toctrees:

• Getting Started — Getting Started, Development Guide
• Project Architecture — Comet Plugin Overview, Arrow FFI, JVM Shuffle, Native Shuffle, ANSI Error Propagation
• Adding Functionality — Adding a New Operator, Expression, Spark Version
• Testing — Comet SQL Tests, Spark SQL Tests, Iceberg Spark Tests
• Debugging and Performance — Debugging Guide, Benchmarking Guide, Profiling, Tracing
• Reference — Supported Spark Expressions, Supported Spark Configurations
• Project Mechanics — Bug Triage, Release Process, Roadmap, GitHub and Issue Tracker

No page renames, no file moves, no content rewrites in the page bodies. The link labels in the toctree directives are unchanged from the originals.

docs/source/_templates/docs-sidebar.html — switch from the bare Sphinx toctree() macro to pydata-sphinx-theme's generate_toctree_html() with a section-relative startdepth so captions render in the sidebar:

• user-guide/latest/* starts at depth 2 (under user-guide/index → latest/index)
• contributor-guide/* starts at depth 1 (under contributor-guide/index)
• Root, search, genindex and other orphan pages fall back to startdepth=0 (global TOC, same as before)

The fallback uses pydata's suppress_sidebar_toctree(startdepth=..., includehidden=True) so pages without a captioned ancestor still render their sidebar instead of erroring.

Out of scope for this PR (deferred from the issue):

• Moving Building From Source into the contributor guide
• New pages (Glossary, Architecture-in-5, FAQ, Version Support)
• Tighter "how to use this guide" intros on the two index pages
• Applying Establish nomenclature style guide and audit operator names for clarity #4419 nomenclature

How are these changes tested?

Built the docs with sphinx-build -b html source build and verified:

1. Build exits clean (0 new warnings — the 4 existing warnings about user-guide/0.{13,14,15,16}/index are pre-existing and come from versions generated by generate-versions.py at build time).
2. Each user-guide/latest page renders the 5 new captions in the sidebar in order; each contributor-guide page renders the 7 new captions in order.
3. Prev/next ordering follows the new section order: e.g. installation → configs → datasources → … → kubernetes → source → Contributor Guide and contributing → development → plugin_overview → … → release_process → roadmap.
4. Root index, search, genindex, changelog, about, asf, and the user-guide/index version selector still render correctly with their existing captions.