[codex] bump Jido deps to 2.1 and align zoi

[mikehostetler]

Summary

Updates the package from the Jido 2.0 release candidates to the stable 2.1 line, based on the focused work from #9, and adds the missing zoi requirement bump needed to resolve dependencies cleanly.

What Changed

• bump jido to ~> 2.1
• bump jido_action to ~> 2.1
• move jido_signal from ~> 2.0.0-rc to stable ~> 2.0
• add {:libgraph, "~> 0.16.0", override: true} to resolve the multigraph / libgraph conflict
• update the telemetry integration test to expect Jido.Action.Error.InternalError
• bump zoi to ~> 0.17 so the new Jido stack resolves without dependency conflicts

Why

PR #9 was the cleanest of the open dependency-bump proposals, but it could not be validated as-is because jido_signal now requires zoi ~> 0.17.1 while the project still pinned zoi ~> 0.14. This patch keeps the dependency bump focused and fixes that resolver mismatch.

Validation

• mix deps.get
• MIX_ENV=test mix test
• mix quality

Closes #8

[mikehostetler]

Validated locally and on CI:

• mix deps.get, MIX_ENV=test mix test, and mix quality all pass locally with the Jido 2.1 bump plus zoi ~> 0.17
• GitHub CI passes all checks except CI/Lint / Lint
• The failing step is mix hex.publish --dry-run, which rejects the current {:libgraph, "~> 0.16.0", override: true} workaround with:
  • ** (Mix) Can't build package with overridden dependency libgraph, remove �override: true�

I also tested replacing that override with {:libgraph, "~> 0.16.1-mg.1", hex: :multigraph} locally, but Mix still resolves both packages and errors with:

• Multiple packages resolved with the same OTP application name of "libgraph": libgraph 0.16.0, multigraph 0.16.1-mg.1

So the remaining blocker appears to be ecosystem-level: reactor still wants hexpm/libgraph, while jido_action wants hexpm/multigraph under the same OTP app name. This PR is therefore a useful proof branch, but it is not publishable to Hex in its current form.