fix(section): honour a per-block padding override

[markdumay]

The problem

cta and panels both declare padding in their bookshop schema, so authors set it and reasonably expect the section's vertical padding to change. It never did.

utilities/section.html derived padding solely from utilities/GetPadding.html, which reads site parameters, and no block passed anything through. The key was accepted, range-validated, and then dropped on the floor.

Measured on a real site before this change — four cta blocks carrying padding: 4 and seven carrying nothing, all identical:

padding: 4   ->  container-fluid px-3 px-md-5 py-5
(none)       ->  container-fluid px-3 px-md-5 py-5
padding: 0   ->  container-fluid px-3 px-md-5 py-5

After

padding: 4   ->  container-fluid px-3 px-md-5 py-4
(none)       ->  container-fluid px-3 px-md-5 py-5     <- site default, unchanged
padding: 0   ->  container-fluid px-3 px-md-5 py-0

The horizontal axis is byte-identical in every case.

The empty default: is load-bearing

padding inherits the shared definition in mod-utils data/structures/_arguments.yml:

padding:
  type: int
  optional: true
  default: 3
  options: { min: 0, max: 5 }

Left alone, default: 3 arrives whenever a caller omits the argument — so "not supplied" is indistinguishable from "supplied", and every section that does not set one gets silently repadded. An earlier revision of this PR hit exactly that: unset cta blocks dropped from py-5 to py-3.

data/structures/section.yml therefore redeclares padding with an empty default:. ArgsSchema.html merges the local node over the global one (merge $def $val), so the empty value wins and a real nil is restored:

padding: 4   ->  $args.padding = 4     (uint64)
(none)       ->  $args.padding = <nil>

The inherited options survive that merge, so validation is unchanged — padding: 9 still fails before rendering:

ERROR partial [.../cta.hugo.html] - Invalid arguments: zz-pad/_index.md
      [cta] argument 'padding': value '9' out of range [-, 5]

This keeps the author-facing name and needs no new argument.

The nil comparison in section.html is deliberate for the same family of reason: with or | default would discard a legitimate padding: 0.

Scope

Vertical axis only. x is the page gutter — shared by every section and also feeding the outer container — so a per-block override there would misalign that section against its neighbours. Any mobile y override is dropped alongside the value it derives from, so the override then applies at every width.

Two blocks only. cta and panels are where padding was purely inert. approach, articles, cards, team and testimonials also declare it, but forward it to inner partials (card group, section title) where it already means something. Making it drive section padding as well would change existing sites — a semantic decision rather than a bug fix, so it is left alone and noted here instead.

Checks

• All four cases exercised on a real 1119-page site: 4 -> py-4, 0 -> py-0, omitted -> py-5, 9 -> rejected.
• Verified the omitted case matches a pre-change baseline build exactly, so no existing section moves.
• pnpm test (exampleSite build) passes: 40 pages, no errors.

[markdumay]

🎉 This PR is included in version 2.3.3 🎉

The release is available on GitHub release

Your semantic-release bot 📦🚀