docs: Why H3 rationale page + iSamples Explorer layout/code-display cleanup

[rdhyee]

This PR bundles three unrelated cleanups on the same branch:

1. Why H3? rationale page (tutorials/why_h3.qmd)

Documents why iSamples uses H3 — and why resolutions 4/6/8 specifically — so newcomers don't have to reverse-engineer the choice from the H3 tier files. Motivated by Raymond's observation that H3 has no standalone Wikipedia article, only a mention in Discrete global grid. Without that fallback, the site needs to make the design choice legible itself.

What the page covers:

• What H3 is, in two sentences, with authoritative links (h3geo.org, GitHub, Uber Eng blog, Sahr 2008)
• Why hexagons over squares (uniform neighbor distance, isotropic direction sampling)
• Why H3 over S2 or geohash (DuckDB integration, 64-bit integer index, dateline behavior)
• Why resolutions 4 / 6 / 8 specifically (cell size × altitude band × parquet payload budget — adding 5 / 7 / 9 ~triples on-the-wire bytes for a barely-perceptible smoothness win)
• Two query patterns (region aggregate via tier file; per-cell filter via wide h3_res6 column)
• What we'd revisit (dominant-source-per-cell color encoding can hide near-equal second sources — Eric Kansa flagged this in Dec 2025)

Wiring:

• Added under "Architecture & Vocabularies" in both the navbar dropdown and the docked sidebar, alongside the existing "Technical: Narrow vs Wide" page (matches that precedent)
• Cross-linked from data.qmd §4 (the H3 tier breakpoints table) and from how-to-use.qmd's H3 helpers section

2. Explorer: hide all OJS code (tutorials/isamples_explorer.qmd)

Replaces code-fold: true with echo: false at the format level so source code blocks no longer appear at all (previously shown behind a "Code" toggle that took visual space). Removes per-cell //| code-fold directives that were overriding the format setting, and drops the now-empty "Setup" and "Database & Queries" section headings (their cells produced no rendered output).

3. Explorer: layout A — globe in right column, table full-width

Moves the Cesium globe into the right column of the filters grid so it's visible immediately (was below the fold under its own ## Globe View heading). Drops the viewMode radio (globe / list / table) — globe is now always visible top-right and the results table renders full width below the grid as ### Results. Removes the now-unused list view, viewMode-keyed display toggles on the cesiumContainer/legend, and the v2-only deferred-mount branch in the viewer cell.

Test plan

• quarto render tutorials/why_h3.qmd succeeds
• quarto render data.qmd and quarto render how-to-use.qmd succeed (cross-link renders correctly)
• Anchor #h3-tier-breakpoints-for-map-authors exists in docs/data.html (target of the back-link from why_h3)
• quarto render tutorials/isamples_explorer.qmd succeeds; rendered HTML has zero visible code blocks; H2 sections are Search & Filters → Sample Card → Debug Info; cesiumContainer renders inside the Search & Filters grid right column; full-width results table appears below
• Deployed to rdhyee fork preview (https://rdhyee.github.io/isamplesorg.github.io/tutorials/isamples_explorer.html?sources=OPENCONTEXT) — visual verification by Raymond
• Reviewer eyeballs claims in §"Why H3 over S2 or geohash" — I assert we did not benchmark S2 head-to-head; if that's wrong, we should soften the section

🤖 Generated with Claude Code

[rdhyee]

Review finding: the new Why H3 page says the wide parquet carries h3_res4/6/8, and the example queries https://data.isamples.org/current/wide.parquet with WHERE h3_res6 = .... That alias currently points to the 202604 wide parquet, which does not have H3 columns, so the copy-paste query fails with Referenced column "h3_res6" not found.

The H3 columns are in isamples_202601_wide_h3.parquet per SERIALIZATIONS.md. Please update the prose and the example to point to that H3-augmented wide file instead of current/wide.parquet.

Checks I ran:

• quarto render tutorials/why_h3.qmd
• quarto render data.qmd
• quarto render how-to-use.qmd
• git diff --check
• DuckDB schema/query checks against local wide and wide_h3 parquet files