fix: clarify mark-read target types

[shaypal5]

Summary

• Accept documented and intuitive mark-read target aliases, including dm, direct-message, and direct-conversation for conversation.
• Return actionable API validation details with valid target types and accepted aliases.
• Update CLI help, API docs, quickstart docs, and schema metadata so callers do not have to inspect source.
• Add API and CLI regression tests for alias normalization and invalid target-type errors.

Self-review fixes applied

Self-review prompt used: "Review the PR yourself, and take an outsider's view: Pretend you're a senior dev and you really don't like the way this PR is implemented; be brutally honest, and aim to make this PR much better with you feedback."

Finding: the initial CLI implementation normalized mark-read target type inside the request payload, after resolving the agent id. That meant a pure local usage error could still require API/token setup before reporting the invalid target type.

Recommended action/fix: normalize and validate the target type before resolveAgentId(...) or any API request. Add a focused CLI test that runs with no API environment and expects the actionable invalid-target response.

Applied: mark-read now computes targetType before agent resolution, and tests/cli.test.ts covers the no-API-env invalid target path.

Validation

• npm test -- --run tests/api-auth.test.ts tests/cli.test.ts
• npm test
• npm run check
• git diff --check
• npm run build

Closes #84