Add experimental era ranges: eon-like spans as type-level lists

[carbolymer]

Context

This PR prototypes era ranges: a generic replacement for the hand-written eon machinery, where an era span is a type-level list sliced from a single timeline instead of a bespoke GADT.

Onwards AlonzoEra         -- Alonzo and everything after it
ShelleyEra :-: BabbageEra -- the closed interval from Shelley to Babbage

The motivation: each eon costs a ~110-150 line module (GADT, Eon/Convert instances, a ~33-item constraint synonym, a CPS reifier, an Is*BasedEra class), adding an era touches all of them, and every constraint bundle except alonzoEraOnwardsConstraints currently ends in a runtime error "TODO Dijkstra".

What this adds, under Cardano.Api.Experimental.Era.Range* (purely additive - no existing eon code is changed):

• EraIn range era - one generic membership witness replacing per-range GADTs, with an Eon instance so existing combinators (forEraInEon, Featured, ...) work with any range.
• All/withRange/rangeDict - range constraints are derived by GHC from the per-era instances instead of hand-listed; an era lacking support becomes a compile error at the use site instead of a runtime error.
• splitRange - total, nestable boundary dispatch replacing the caseShelleyToXOrYEraOnwards family (which errors on Dijkstra and cannot be nested).
• SomeEraIn range - a vary-based existential (Vary of CardanoEra singletons) with O(1) morph widening between ranges.
• Compile-time canaries proving the timeline matches the ledger's PreviousEra chain (timelineMatchesLedger) and that SupportedEras tracks the experimental Era GADT (supportedErasAligned).
• A fix for TestEquality CardanoEra, which was missing the DijkstraEra case (Dijkstra silently tested unequal to itself).

Constraint bundles follow one idiom, with the constraint list written exactly once: a plain alias as the single source of truth (used applied in signatures) plus a two-line class/instance eta-expansion making it passable to withRange unapplied.

Full rationale, trade-offs (including the honest overlap with vary's internals) and migration story are in era-ranges-with-vary-design.md at the repo root.

New dependencies: vary (Hackage, within the pinned index-state) and constraints (already in the build-plan closure).

How to trust this PR

• The change is additive: apart from the one-line TestEquality fix and cabal wiring, no existing module is touched.

• Test.Cardano.Api.Experimental.EraRange doubles as the usage showcase: lower bound (ConwayEraTxCert incl. Dijkstra), upper bound (ShelleyEraTxCert, capped at Conway by the class itself), subset narrowing and nested three-way dispatch via splitRange, membership checks (including the TestEquality regression), and a vary round-trip. Run them with:

  cabal run cardano-api:test:cardano-api-test -- -p "EraRange"
  

• The two canaries make the central claims compile-time checked: if the timeline drifted from the ledger's era chain, or SupportedEras from the experimental Era GADT, the library would not build.

• Must-NOT-compile counterexamples (requesting ShelleyEraTxCert for a Dijkstra-containing range; an era outside the timeline) are documented next to the tests.

• Library and tests build warning-free under -Wall -Wredundant-constraints -Wunused-packages; hlint reports no hints.

Checklist

• Commit sequence broadly makes sense and commits have useful messages
• New tests are added if needed and existing tests are updated. See Running tests for more details
• Self-reviewed the diff
• Changelog fragment added in .changes/

[Copilot]

Pull request overview

This PR introduces an experimental “era ranges” abstraction that models era spans as type-level lists sliced from a single timeline, providing generic membership witnesses, derived range-wide constraints, and total/nestable boundary dispatch. The intent is to replace the current hand-written “eon” boilerplate with a smaller generic core while also fixing a correctness bug in TestEquality CardanoEra for DijkstraEra.

Changes:

• Add Cardano.Api.Experimental.Era.Range (and .Some) implementing EraIn, KnownEras, All/withRange/rangeDict, and total splitRange, plus a vary-based existential SomeEraIn.
• Add a new test module Test.Cardano.Api.Experimental.EraRange covering lower/upper bounded ranges, subset narrowing via splitRange, nested splits, membership checks, and a vary round-trip.
• Fix TestEquality CardanoEra by adding the missing DijkstraEra reflexive case; wire new modules and dependencies (vary, constraints) into the cabal file and add a changelog fragment.

Reviewed changes

Copilot reviewed 9 out of 9 changed files in this pull request and generated no comments.

Show a summary per file

File	Description
era-ranges-with-vary-design.md	Design/rationale document describing the era-range approach and migration story.
cardano-api/src/Cardano/Api/Experimental/Era/Range.hs	Public experimental API surface for type-level era ranges and associated operations.
cardano-api/src/Cardano/Api/Experimental/Era/Range/Internal.hs	Core implementation: timeline, slicing families, EraIn, KnownEras, All/withRange, and splitRange, plus compile-time canaries.
cardano-api/src/Cardano/Api/Experimental/Era/Range/Some.hs	vary-backed existential SomeEraIn utilities (injectEra, withSomeEra, onEra, isEra).
cardano-api/src/Cardano/Api/Era/Internal/Core.hs	Bugfix: TestEquality CardanoEra now handles DijkstraEra.
cardano-api/test/cardano-api-test/Test/Cardano/Api/Experimental/EraRange.hs	New Hedgehog properties exercising the new range machinery end-to-end.
cardano-api/test/cardano-api-test/cardano-api-test.hs	Registers the new EraRange test group.
cardano-api/cardano-api.cabal	Exposes the new experimental modules; adds constraints and vary; adds internal module to other-modules; adds test module.
.changes/20260716_cardano_api_experimental_era_ranges.yml	Changelog fragment classifying this as feature + bugfix with a detailed description.