Audit Deeplink workflows to improve UX of handoff to GL

[d13]

From Ramin in Slack:

Louis and I just went through the experience opening deep links into the graph (with both disk path and remote url) and it's not great at all. Feels like there are several regressions, some quite concerning. I'm going to be writing up the bugs/pain points in the flows during the experience and, if we have room in our next cycle, would like them included, especially since these are entrypoints into our product from many different surfaces

To list some of the most concerning ones:

• Supplying a disk path opens the repo "silently", without adding it to the workspace, and without prompting how the user wants to open it, and shows it in the graph. If you already have a repo open, this is even more confusing because the header of the graph view still shows the other repo's name.
• Supplying a remote url, if you don't have the remote we prompt you to add it, but after adding it we get "Unable to resolve link" and the link totally dies there. It should go through successfully after you add the remote.
• We are failing to detect the repo if you supply a remote URL in a different format (HTTPS/SSH) than the way you cloned the same remote. We should try a bit harder to match the remotes than exact-matching the remote urls