feat(templates): declare which presets can be handed a runtime module

[DemchaAV]

Stacked on #580 (base feat/cv-module-sections), which is itself stacked on #579. Retarget down the chain as each merges.

Why

The previous PR made a CV section's shape a runtime value. That is only worth anything if the template you hand it to renders it — and not every preset can promise that. Several compose a fixed set of modules and find each by matching headings against private keyword lists, so a section they do not recognise never reaches a renderer at all. The CV still comes out, minus a section, looking finished. Nothing about that failure is visible at the point it happens; it surfaces weeks later as a job that was never listed.

So the question "can I give this template a module I invented?" needs an answer that is checked rather than assumed. And a second question follows it: if a preset can render the module, can it render it in its own style rather than the canonical one — the limitation #580 documented and left open.

What changed

• ModularCvTemplate is the promise; CvTemplates.modular() is the list to offer. The registry also answers byId, all, ids, and recommendedMargin (the margin a preset was designed at, which a caller needs while building the session, before it has a template). Picking a template at runtime stops being a map kept by hand in each consumer — the one this model exists for keeps exactly that map in another repository, where nothing tells it a preset shipped.
• Declaring the interface costs something. ModularCvTemplateFidelityTest enumerates CvTemplates.modular() and every CvKind — not lists of its own — and renders: every kind, a section this catalogue has no name for, a heading in a script no keyword list contains, and a heading that does match one. Seven presets qualify. ClassicSerif does not — it draws any shape it is given, but only gives itself the sections it recognises. Finding that out is what the gate is for; it kept its kit and lost the claim.
• The promise names its scope: Slot.MAIN. Every shipped preset composes one main column, so a section placed in Slot.SIDEBAR is dropped — by these templates as by every other, which is what CvDocument has always documented. Saying so in the contract, and pinning it in the gate, is the difference between a contract and a slogan; when slots go live that test goes red and the promise moves with the code.
• CvRenderKit is the three shapes a section body reduces to — a paragraph, a label/value row, a timeline entry — and a template hands back the kit it draws them with. The lowering from CvItem stays shared: what a linked title looks like, which fields a kind reads, what an empty description does to a trailing colon are the model's decisions and must not be re-made sixteen times. It is also the shape the presets already had — their private renderers take a CvEntry or a CvRow today. BlueBanner, ClassicSerif and EditorialBlue now draw modules with their own entry and project shapes; the rest return CvRenderKit.defaults(), and every one of them passes its kit to the dispatcher so overriding it later actually does something.
• EditorialBlue stops renaming a module's heading. Its keyword vocabulary turned "Certifications & Awards" into EDUCATION — the one thing the promise says cannot happen. The canonical sections keep the rename that gives the preset its voice.
• CvTemplatesCoverageTest derives the catalogue from the presets package rather than trusting it, so a preset that ships unregistered fails the build instead of being invisible to every caller that looks one up by id. No preset list in this repo had such a guard.

Verification

• Full reactor gate — ./mvnw -B -ntp clean verify -pl :graph-compose-core,:graph-compose-render-pdf,:graph-compose-render-docx,:graph-compose-render-pptx,:graph-compose-templates,:graph-compose-testing,:graph-compose-qa,:graph-compose-coverage -am → BUILD SUCCESS, qa 821 tests (+14). The 16 CV visual baselines are untouched: the three-argument SectionDispatcher.renderBody delegates with defaults(), whose methods are the same static renderers, and the kit-passing presets reach the four-argument form only from their module branch.
• Fidelity gate, 43 cases across the seven modular templates. Proven non-vacuous: restoring EditorialBlue's heading rename turns aHeadingThePresetHasAWordForIsStillTheAuthorsHeading red for editorial-blue and nothing else.
• Registry guard, 6 cases, proven fail-closed: removing Panel from the catalogue fails everyPresetInThePackageIsInTheCatalogue with the directory named in the message.
• japicmp (both baselines, both modules) → BUILD SUCCESS — the new types are additions and SectionDispatcher gained an overload rather than changing one. javadoc:javadoc → 0 warnings.
• Two cases that duplicated the new gate were moved out of ModuleSectionKindCoverageTest rather than copied, so the 16×6 documents are rendered once.

Lane: templates (api + components + registry + nine presets).

Limits:

• Eight presets still guard their module slots on the section's Java type and are not modular; that, SectionRole-driven routing, and live slots are the next PR.
• EditorialBlue's kit draws entries in its experience styling: it picks between experience and education by sniffing a heading, which is what a module carries a role to avoid, and the kit is not handed that role yet.
• A description line the author asked to bullet is drawn by the shared primitives rather than the kit — a bullet glyph is the shape the model was asked for, not a style a template gets a say in.