fzf#run() based on the results of fzf#run() + function

- Category
- [X] Question
- [ ] Bug
- [ ] Suggestion
- OS
- [X] Linux
- [ ] macOS
- [ ] Windows
- [ ] Etc.
- Vim
- [ ] Vim
- [X] Neovim

I'm attempting to use a fzf#run to determine what directory to select, and then launch a GFiles. The actual workflow is to select from a list of git repos, then cd to the selected repo and use GFiles to pick an initial file to open. (This is intended to approximate the emacs/helm/projectile command `helm-projectile-switch-project`.)

I understand that my current implementation, using a sink to launch another FZF completion doesn't work, but I can't figure out if what I want is actually supported without some sort of sleep-style hack. Is there a way to accomplish effectively piping the output one fzf into another?

```
" Current implementation. The problem here is that we run CdGFiles inside of an fzf#run()
" :CdGFiles will cd to and then run GFiles
function! CdGFilesImpl(directory)
exec 'cd' a:directory
GFiles
endfunction
command! -bang -nargs=1 CdGFiles call CdGFilesImpl()

function! ProjectileProject()
call fzf#run(fzf#wrap({'source': 'bash -c "find ~/code -maxdepth 2 -type d -name \"*.git\" -exec dirname {} \; | sort"',
\ 'sink': 'CdGFiles'}))
endfunction
nnoremap pp :call ProjectileProject()
```

Start by tracing the fzf#run(), fzf#wrap(), and sink handling used by the CdGFiles command in the Vim integration. Determine whether one fzf invocation can launch another while preserving the selected directory, and define completion by either a supported nested workflow or clear documentation of the limitation.

Written by the indexing model from the issue text.