fix: do not cleanup api and repo.lock files

[hacdias]

Previously we were cleaning up the api and repo.lock files because the shutdown on Windows was not working correctly on the IPFS daemon. It seems the issue was already resolved (see ipfs/kubo#3061).

This PR removes that code and this should also fix #689 and fix #688.

@lidel can you check this out, please?

[hacdias]

I also removed our path definition and now it uses ipfsd-ctl default path which is the IPFS_PATH variable if present or $HOME/.ipfs otherwise.

@lidel we now have a small issue: if the repo has a api file and doesn't have a repo.lock file, js-ipfsd-ctl will try to connect to the API and will fail if the daemon isn't running. Do you think we should catch the ECONNREFUSED error and remove the api file?

[lidel]

Works as expected and improves things a lot for me, thanks! 👍
(I was able to use external node from Docker)

Do you think we should catch the ECONNREFUSED error and remove the api file?

I strongly believe in "no surprises" when it comes to desktop apps, so we should aim at not removing anything from $IPFS_PATH/ without explicit consent from user.

For now we should just detect ECONNREFUSED and display human-readable error window that remote node defined in $IPFS_PATH/api is offline and add "Retry" and "Quit" buttons (leaving it up to technical user to start node or edit the address in file).

In the long run (future PR), we probably want to make this built-in and super nice, wrote some ideas in: #713

[hacdias]

Thanks @lidel. I'll create the dialog on a different PR!