Use github_remote_list() instead of github_remotes() when GitHub not consulted

[hadley]

To reduce some of the usage of github_remotes() where it's not strictly necessary.

[jennybc]

My main reaction is: surely I had some reason for using github_remotes(github_get = FALSE) instead of github_remote_list() (?). Because obviously I had both fully uploaded into my brain when I did this. I'm just getting some Chesterton's Fence vibes. It's hard to know what will happen since none of this is under test.

[hadley]

new_*() are only used for tests, so I think that's safe.

[jennybc]

new_*() are only used for tests, so I think that's safe.

I do have some memory that the whole reason I made the new_*() functions this way was so that I could have authentically "shaped" github remote info.

I.e. sure it is github remote info that is fake, but I make it in such a way that it's guaranteed to have the same "shape" as when we actually do call github. I think this can't be detected today (meaning: tests aren't failing), because the assignments we go on to make happen to create all the right fields. The fields that are no longer there due to the switch from github_remotes(github_get = FALSE) to github_remote_list() all get created now, instead of just modified.

[jennybc]

So I think this implies that the PR should probably only propose a change in pr_create()?

[hadley]

Giving up on this approach