feat(ruby): release runbook

[TheVaultdweller13]

🚪 Why?

This wires the Ruby SDK into the same release machinery TypeScript and Python
already use — one quarterly runbook, one endpoint reference, one automated
versioning-and-publishing pipeline for the three SDKs — ahead of the next
dated API version in October.

🔑 What?

• Runbook (.agents/skills/release-sdk/SKILL.md) — adds the "Regenerate
  Ruby" step to the quarterly regeneration guide. Regenerating reuses the
  version already in version.rb, so the step needs no version revert.
• Skill reference — sdk-methods.md gains a Ruby column for all 550
  endpoints. A new ruby/scripts/skill_methods.rb extracts each call by
  reflecting on the loaded gem (naming stays owned by normalize_oas.rb), so
  the tables always match what the gem actually exposes; if the Ruby toolchain
  is unavailable, generation fails loudly rather than emitting a partial
  table. The Ruby pipeline now refreshes the skill on every regeneration, the
  same anti-drift loop the TS/Python release scripts run.
• release-please + publishing — registers the ruby package (config +
  manifest), hands version ownership to release-please (generate_sdk.rb now
  reads version.rb instead of computing bumps; --set-version remains for
  the beta workflow), and wires publish.yaml to a new publish-to-rubygems
  composite action (idempotent: skips if the version is already on the
  registry).
• Docs — CLAUDE.md, RELEASING.md, ruby/DEVELOPMENT.md and the
  consumer skill now cover the three SDKs.

Merging this publishes nothing: release-please opens a Ruby Release PR once
features land, and only merging that publishes — which also requires the
RUBYGEMS_API_KEY secret (requested separately). Out of scope: the Ruby job
in beta-publish.yaml (small follow-up touching beta-detect.mjs).

[jorgeteixe]

Overall looks fine. Some nitpicks, up to you:

• I think the Gemfile.lock will be drifting in every release... maybe its worth to untrack it or make a commit back? I am not sure of the solution but you get the idea.
• We should make sure the first release is 2.x.x!
• sdk-methods.md's Ruby cells show no arguments

[TheVaultdweller13]

@jorgeteixe

1. Same as uv.lock: gem build reads the version from the gemspec, not the
   lockfile, so a stale line there can't affect what gets published (checked it. Bumping version.rb with a stale lockfile still builds the right version).
   Extended the RELEASING.md note to mention it.
   Left it tracked because it pins RuboCop: with NewCops: enable, an unpinned version would break CI whenever RuboCop ships new cops.

2. Guaranteed: the manifest is seeded at 2.0.0, so any computed release is 2.x or higher. Note the first published version will be the next bump (2.1.0). Documented in RELEASING.md.

3. Good catch. Ruby cells now carry the real required arguments, read by reflection from the gem.

[jacobobq]

Looking good!