This fixes the Windows regression I mentioned on IRC that was flagged by our
CI. It turned out that the bug is not Windows specific at all, it's just
that on Windows setCurrentDirectory with a http URL fails because that is
not a valid path name. Calling error in such cases lets us catch this when
testing on Linux, too.

A more principled solution would be to parse arguments that denote a repo
location immediately in the frontend and use the parsed form in the rest of
the code instead of the raw string.

3 patches for repository http://darcs.net/screened:

patch d7a11386d72441f3a00aa876360bdd8de21f3b4e
Author: Ben Franksen <ben.franksen@online.de>
Date:   Thu Jul  8 10:33:14 CEST 2021
  * setCurrentDirectory: call error if argument is a remote URL

patch 46c59a13b975e80ed869082745f09b423a4febee
Author: Ben Franksen <ben.franksen@online.de>
Date:   Sun Feb 28 11:44:48 CET 2021
  * fix in checkSuspendedStatus and maybeDisplaySuspendedStatus

  We must not try to access either of the rebase patch files if the repo
  location is a remote URL.

patch b2775f40b6521e2df3f555361b4ff9de6f4fc25c
Author: Ben Franksen <ben.franksen@online.de>
Date:   Thu Jul  8 20:37:14 CEST 2021
  * HasCallStack: withCurrentDirectory, withRepoDir

Attachments

• patch-preview.txt
• setcurrentdirectory_-call-error-if-argument-is-a-remote-url.dpatch

Looks good - one question:

>   * fix in checkSuspendedStatus and maybeDisplaySuspendedStatus

>  We must not try to access either of the rebase patch files if the repo
>  location is a remote URL.

Are there scenarios in which we call these functions for a remote repo?
Maybe those could be explained in a comment on the `isValidLocalPath`
guards if so.