feat: register Claude Fable 5.1 model in registry

[staging-supernova-dx-appf-io]

What this does

Registers Claude Fable 5.1 (released by Anthropic on 2026-09-01, announced 2026-08-31) in ruby_llm's model and alias registries, following the same pattern used previously for Claude Opus 5 and Claude Fable 5. This adds claude-fable-5-1 entries across all providers/regions that claude-fable-5 already has (anthropic, Bedrock eu/global/us, vertexai), with matching modalities/capabilities, context window, max output tokens, and reasoning effort options, plus a corresponding alias entry. Pricing (input/output) is unchanged from Fable 5; cache-write pricing is also unchanged. Cache-read pricing reflects a reported cut sourced from AWS/Microsoft partner blog posts that has not been independently verified against Anthropic's live pricing page — this is deliberately flagged in code comments, following the repo's existing precedent for the GPT-5.6 cache-write pricing gap. A new spec file (spec/ruby_llm/models_fable_5_1_spec.rb) verifies registration, pricing (including the flagged figures), reasoning effort options, alias resolution, and Bedrock regional lookup.

Claude Mythos 5.1 (a fewer-safeguards variant released alongside Fable 5.1) is explicitly excluded from this change, mirroring the precedent that only Fable 5 (not Mythos 5) was registered before.

This PR also includes two incidental CI-unblocking fixes needed to get this branch green, unrelated to the original model-registration ask:

• Fixed 14 RuboCop Style/DirectiveScope offenses by converting single-line # rubocop:disable comments to # rubocop:disable-next across several lib and spec files (comment-only, no logic change).
• Fixed a JRuby CI failure caused by the archspec dev gem lacking a JRuby-compatible native extension, by adding platform: :mri to its Gemfile line (mirroring the existing async gem's platform constraint) and regenerating the four Appraisal-generated gemfiles (rails_7.1, 7.2, 8.0, 8.1) to propagate the change.

Type of change

• Bug fix
• New feature
• Breaking change
• Documentation
• Performance improvement

Scope check

• I read the Contributing Guide
• This aligns with RubyLLM's focus on LLM communication
• This isn't application-specific logic that belongs in user code
• This benefits most users, not just my specific use case

Required for new features

• I opened an issue before writing code and received maintainer approval
• Linked issue: #___

No prior approved issue was identified for this specific change during this session; flagging per the template's note that PRs for new features without a prior approved issue may be closed.

PRs for new features or enhancements without a prior approved issue will be closed.

Quality check

• I ran overcommit --install and all hooks pass
• I tested my changes thoroughly
  • For provider changes: Re-recorded VCR cassettes with bundle exec rake vcr:record[provider_name]
  • All tests pass: bundle exec rspec
• I updated documentation if needed
• I didn't modify auto-generated files manually (models.json, aliases.json)

Note: models.json and aliases.json were edited directly and manually as part of this change, which conflicts with the last item above. This mirrors the precedent set by prior Opus 5 / Fable 5 registration commits, where direct manual edits to these files were the established approach. Only the new Fable 5.1 spec, the existing model-registry validation spec, and a targeted run of models_spec.rb were run — not the full bundle exec rspec suite — per explicit instruction to avoid running the full suite. RuboCop was run on changed files after autocorrect, with zero offenses remaining.

AI-generated code

• I used AI tools to help write this code
• I have reviewed and understand all generated code (required if above is checked)

This entire change was produced by an AI coding session.

API changes

• Breaking change
• New public methods/classes
• Changed method signatures
• No API changes

This is registry data (model/alias JSON) plus a Gemfile/Appraisal-gemfile platform constraint and comment-style RuboCop fixes — no code API changes.

---

In Supernova
Chat: open

[staging-supernova-dx-appf-io]

🤖 Supernova Code Review — View trace

♻️ Re-run this automation

— via automation "Code Review Sonnet 5 Medium" (run), installed by Sam Boland (sam.boland@appfolio.com)

[staging-supernova-dx-appf-io]

Reviewed the diff against the existing claude-fable-5 entries in models.json/aliases.json line-by-line (base anthropic, EU/global/US bedrock, vertexai). All fields mirror the precedent structure exactly, and the cache_read_input_per_million price cut is scaled proportionally and consistently across regions (base 1 → 0.25, EU 1.1 → 0.275, matching the 4x reduction ratio). The new spec file correctly flags the unverified cache-read pricing per repo convention. No actionable bugs, logic errors, or maintainability issues found — no inline comments.

♻️ Re-run this automation

— via automation "Code Review Sonnet 5 Medium" (run), installed by Sam Boland (sam.boland@appfolio.com)