Improve oxide-instance builder documentation

Summary

Several areas of the oxide-instance builder documentation are missing or underdeveloped. This issue tracks the documentation improvements needed to make the plugin more approachable and production-ready for users.

Items

Authentication guidance
Document the security tradeoffs between inline host/token fields versus environment variables (OXIDE_HOST, OXIDE_TOKEN). Inline credentials risk being committed to version control; env vars are preferred for CI/CD pipelines. Add guidance on token scoping and rotation.

Cleanup behavior on failure
Explicitly document what resources are and are not cleaned up when a build fails. Users need to know whether a failed build can leave behind a running instance, an orphaned disk, or a partial image — and if so, how to find and clean up those resources.

Timeout defaults
Document the current hardcoded timeout values for instance boot, SSH readiness, and image creation. Even if these are not yet user-configurable, knowing the defaults helps users understand why a build is hanging and set expectations for long-running provisioners.

No documentation file or test is named. Start by locating the oxide-instance builder documentation and the implementation of authentication, failure cleanup, and timeout handling. Done means the docs explain credential choices and token practices, resource cleanup after failures, and the current boot, SSH-readiness, and image-creation timeout defaults.