Fix issue with haddock generation in CI

[palas]

Changelog

- description: |
    Fixed haddock generation bug by using GHC 9.14
  type:
    - maintenance
  projects:
    - cardano-api

Context

The GitHub Pages haddock generation workflow is currently broken due to a GHC 9.12.2 panic:

panic! (the 'impossible' happened)
  GHC version 9.12.2:
	tyConStupidTheta

AllLedgerPeers

This PR works around the issue by:

1. Adding a haddockCompiler = "ghc914" variable in flake.nix and including it in the flake variants so that a .#ghc914 dev shell is available on x86_64-linux.
2. Updating .github/workflows/github-page.yml to use the .#ghc914 dev shell instead of the default (GHC 9.12.2) shell.

How to trust this PR

• The flake change is minimal: a new variable and adding it to the existing variants list.
• The GHA change just adds devShell: .#ghc914 to the use-nix-shell-action step.
• I've triggered the workflow on this branch so you can see the result here.

Checklist

• Commit sequence broadly makes sense and commits have useful messages
• New tests are added if needed and existing tests are updated
• Self-reviewed the diff

[Copilot]

Pull request overview

Work around a GHC 9.12.2 Haddock panic in the GitHub Pages docs workflow by switching Haddock generation to a GHC 9.14 dev shell.

Changes:

• Add a haddockCompiler = "ghc914" flake variable and include it in the Linux compiler variants.
• Update the GitHub Pages workflow to enter nix develop .#ghc914 (via use-nix-shell-action) before building Haddocks.

Reviewed changes

Copilot reviewed 2 out of 2 changed files in this pull request and generated 1 comment.

File	Description
flake.nix	Introduces a GHC 9.14 compiler variant intended for Haddock generation and includes it in the variants set.
.github/workflows/github-page.yml	Switches the docs workflow to use the .#ghc914 dev shell.

[carbolymer]

Suggested change

	devShell: .#ghc914
	devShell: .#haddock

Could we use haddock name for the haddock shell? It'll be a tad more maintainable in case ghc version changes.

[palas]

You have a point in that, if you change the ghc version for that shell, then the pages ci action is going to break silently. But I don't think changing the name to haddock is a good idea because it kind of suggests the dev shell includes the haddock, which it doesn't.

More arguments against the suggested change:

• It follows the pattern of the existing "stable" shell, which is called ".#ghc967"
• It doesn't have to be exclusive to haddock generation, it is a normal ghc914 shell that could be used for anything really.
• The comment close to where it is defined makes it already explains it is used for haddock generation.

I was initially considering naming the variable something like latest or unstable instead of haddock, but I was afraid that wasn't accurate enough.