fix: use PATH instead of fish_user_paths + tests

[bakotaco]

Summary

Update asdf.fish to modify $PATH directly, rather than using $fish_user_paths. This respects the user's choice between a universal or global $fish_user_paths, addressing an issue where the previous asdf.fish made this decision on behalf of the user.

Modifying $PATH directly was recommended by @faho (long time Fish shell contributor) and avoids assumptions about $fish_user_paths being global or universal, which should be the shell users decision. Note that in fish $PATH cannot be a universal variable (it's simply inherited from the parent process), so it does not have this issue.

This approach also aligns with initialisation scripts from other projects, such as those from brew and pyenv that also set PATH directly, ensuring simplicity.

A side effect is that this also fixes tests and resolves test interdependencies caused by the persistence of the universal variable across runs.

Fixes:

• tests 🙃
• fish_add_path is not persistent fish-shell/fish-shell#9819
• bug: fish path set-up breaks fish_add_path #1629
• fix: use universal scope for fish_user_paths in fish shell integration #1699

[hyperupcall]

Thanks for the fix! Since PATH is not a universal variable, this means that this code must be sourced on every fish shell initialization, right?

Also, it looks like you accidentally commited your Intelij files

[bakotaco]

Since PATH is not a universal variable, this means that this code must be sourced on every fish shell initialization, right?

Yes, that's correct: $PATH is a global variable, so asdf.fish will need to be sourced in every new fish shell session.

looks like you accidentally commited your Intelij files

Oops, removed

[jthegedus]

Thanks for the detailed explanation, cheers!