[vyncint]

Closes #1714.

hello-gpu.md ran cargo oxide doctor straight after installing cargo-oxide, before cargo oxide new. Outside the repository and outside a project, doctor stops with Could not find cuda-oxide workspace or a standalone Cargo.toml, so the guide's second command failed for anyone following it.

This moves the step to after cd my_first_kernel, which is also the order cargo oxide new prints:

  cd my_first_kernel
  cargo oxide doctor
  cargo oxide run

The explanation moves with it. It gains one sentence saying that doctor reads the project's toolchain pin and dependencies, which is why it runs from the project directory.

Checked:

• With a cargo-oxide built from main (538f4839), doctor exits 1 with that error in an empty directory. In the new order (cargo oxide new my_first_kernel, cd, cargo oxide doctor), it exits 0 with 13 checks passing.
• I first hit this following the guide with cargo-oxide 0.3.0 installed from the cuda-oxide/v0.3.0 tag.
• The other doctor steps in the docs already run from a valid directory. The installation chapter and the README run it from the repository checkout, and the debugging chapter assumes an existing project.
• just book builds with warnings as errors.

[copy-pr-bot]

This pull request requires additional validation before any workflows can run on NVIDIA's runners.

Pull request vetters can view their responsibilities here.

Contributors can view more details about this message here.