feat(authz): update OpenFGA model

[casey-brooks]

Summary

• align OpenFGA model with current authorization spec (cluster/org/thread/model, app permissions)
• update model tests to cover new relations and computed permissions

Testing

• buf generate buf.build/agynio/api --path agynio/api/authorization/v1
• go test ./...
• golangci-lint run ./...

Refs #136

[casey-brooks]

Summary

• aligned the OpenFGA model with the current authorization spec (cluster/org/thread/model, app permissions)
• refreshed model test fixtures for updated relations and computed permissions

Testing

• buf generate buf.build/agynio/api --path agynio/api/authorization/v1
• go test ./... (passed: 7, failed: 0, skipped: 0)
• golangci-lint run ./... (no issues)

[noa-lucent]

Model changes look aligned with the authz spec (org/thread/model types + app installation permissions). Inline nits/suggestions left re: minor readability and an extra negative thread-permissions test.

Note: the PR references #136 but that issue number doesn’t exist in this repo (might be cross-repo) — consider linking the correct issue for traceability.

[noa-lucent]

[nit] Follow-up (previous comment may have had formatting issues): consider defining owner before member since member references owner. It reads more naturally and avoids forward-reference when scanning.

[noa-lucent]

[minor] Test coverage: consider adding an explicit negative check for an org member who is not a participant on a thread (e.g. org-member-id on thread:thread-1 should have can_read/can_write/can_add_participant = false). This locks the key invariant that membership alone doesn’t grant thread-level access.