shell setup for better performance

Describe the feature

Currently jup use a simple shim to map node version, it works well but it will cost about 26ms in my environment to run the node (~2.5x time to run native node)

Inspired by nvm and fnm, we can add a hook to shell, update the PATH to the selected node bin on each cd, e.g. on zsh:

# .zshrc
jup-env() {
  (( $+commands[jup] )) || return 0
  local jup_node_store="${JUP_HOME:-$HOME/.cache/jup}/v1/node"

  # drop entries a previous run prepended so each directory re-resolves
  path=(${path:#$jup_node_store/*})

  local node_version node_bin_dir
  node_version=$(jup node -v 2>/dev/null) || return 0
  [[ $node_version == v*.* ]] || return 0
  node_bin_dir="$jup_node_store/${node_version#v}/bin"
  [[ -x $node_bin_dir/node ]] && path=("$node_bin_dir" $path)
}

autoload -Uz add-zsh-hook
add-zsh-hook chpwd jup-env
jup-env

Then the node can be run as fast as native. It may works on other commands.

Can we have a official document or something like jup env --use-on-cd --shell zsh to do that?

Additional information

• Would you be willing to help implement this feature?

Start at the CLI entry point for jup node -v and examine how commands resolve the selected Node version. Compare the proposed zsh chpwd hook with the alternative jup env --use-on-cd --shell zsh, then clarify whether the deliverable is official documentation, a command, or both. Done means the chosen approach updates PATH on directory changes and preserves fast native-like node startup.