Refactor user data bootstrapping logic for Flex

[bcho]

This pull request introduces a major refactor and feature expansion to the Flex node bootstrap configuration, focusing on making the user data generation more flexible, robust, and testable. The key improvements include the adoption of a functional options pattern for configuration, support for multiple CPU architectures, improved validation, and enhanced test coverage.

User Data Generation Refactor and Flexibility

• Refactored the flex.UserData API to use a functional options pattern, allowing callers to specify options like GPU support, Kubernetes version, architecture, and kubeadm config in a more extensible and readable way. This replaces the previous positional-argument approach and makes the API easier to extend in the future. [1] [2]
• Added input validation for supported architectures (amd64, arm64) and non-empty Kubernetes versions, with clear error messages for invalid inputs.
• Updated all call sites to use the new options-based API, improving clarity and maintainability. [1] [2] [3]

Multi-Architecture and Bootstrap Improvements

• Added a new flag (--arch) to the CLI for specifying the CPU architecture, defaulting to amd64. The generated bootstrap script and binary download URLs now dynamically use the selected architecture, enabling support for both amd64 and arm64 nodes. [1] [2] [3]
• The bootstrap script is now generated from a Go template, making it easier to maintain and extend for different architectures and versions. [1] [2] [3]

Testing and Validation Enhancements

• Added comprehensive unit tests for the new UserData options, covering default behavior, architecture selection, invalid inputs, and Kubernetes version handling. This ensures correctness and prevents regressions.

These changes collectively improve the flexibility, safety, and maintainability of Flex node bootstrapping, and lay the groundwork for future enhancements.

References:
[1] [2] [3] [4] [5] [6] [7] [8]