fix: detect host platform in init and clarify remote SSH probe failures

[RobertDeRose]

Summary

• detect the current host platform when system-manager init generates system.nix
• avoid hardcoding x86_64-linux in scaffolds created on non-x86_64-linux hosts
• clarify remote deploy errors so SSH user/auth/connect problems do not look like nix-store PATH failures

Testing

• nix develop --accept-flake-config --command cargo test -p system-manager
• confirmed on an aarch64-linux Lima VM that upstream main generates nixpkgs.hostPlatform = "x86_64-linux";

fixes #469

[eureka-cpu]

Thank you for the contribution and clean individual commits @RobertDeRose 🙂

[eureka-cpu]

The debian container is failing a bit ambiguously, maybe just needs to be retried. All other checks passed.

[RobertDeRose]

@eureka-cpu No problem. I'll see if I can figure out what about this change fails on Debian

[eureka-cpu]

@eureka-cpu No problem. I'll see if I can figure out what about this change fails on Debian

I just re-ran it and it passes, must've been a network error: https://buildbot.numtide.com/#/builders/3075/builds/8

[RobertDeRose]

@jfroche I believe I have addressed all your feedback

[picnoir]

Overall, I think this feature makes sense. But I don't think this is the right approach, it is a bit brittle. Instead of search-replacing x86_64-linux, we should use a generic placeholder and override it to inject the arch-os duo.

That being said, chris approved this, let's not drive you mad with a gazillion back and forth. Let's merge that and I'll fix this in a subsequent commit.

Let's remove the useless tests and merge this.

Also, if this is LLM generated, please review the output before submitting it in the future.

[eureka-cpu]

@picnoir i hadn't reviewed the latest commit, I think originally it looked okay and I agree it is a bit brittle but then we also have to consider that the template must also work for nix flake init --template

[RobertDeRose]

@eureka-cpu I addressed your feedback.
@picnoir this was LLM assisted because I am new to Rust and Nix so I did rely on it to help with the tests and some of the implementation for proper placement. To be honest I did review the generated tests, but I didn't focus on if they actually hit production code which was a mistake and lack of time to ensure they did. I'll be more thorough in the future if I submit another PR.