Implement certain git status operations without spawning git

[joaomoreno]

Like finding HEAD and all refs, etc.

cc @rebornix

[rebornix]

Cool idea! It would also be interesting if we introduce some caching for read operations (similarly to atom/github#700 but probably we already have that?). I'd love to see some performance gain if I batch read operations by purpose.

[lszomoru]

Here is the set of git.exe calls while updating the repository model. Some execute sequentially while some in parallel:

(1) [2022-10-03T11:37:24.163Z] > git status -z -uall [679ms]
(2) [2022-10-03T11:37:24.176Z] > git symbolic-ref --short HEAD [11ms]
(3) [2022-10-03T11:37:24.217Z] > git for-each-ref --format=%(refname)%00%(upstream:short)%00%(objectname)%00%(upstream:track)%00%(upstream:remotename)%00%(upstream:remoteref) refs/heads/main refs/remotes/main [39ms]
(4) [2022-10-03T11:37:24.232Z] > git remote --verbose [12ms]
(5) [2022-10-03T11:37:24.855Z] > git for-each-ref --sort -committerdate --format %(refname) %(objectname) %(*objectname) [637ms]
(6) [2022-10-03T11:37:24.880Z] > git config --get commit.template [11ms]

(1) and (6) are out of scope for the purpose of this work.

[lszomoru]

(2) has been implemented. The rest I will look at next milestone.

[lszomoru]

(4) has been implemented as well.

[lszomoru]

The rest is currently out of scope.

[lszomoru]

Verification steps:

• Open a folder/workspace that contains a git repository with multiple remotes
• Open the "Source Control" viewlet and click the refresh button multiple times
  • Open the git output window, and confirm that there are no executions of git remote --verbose
• Open the Command palette and invoke the "Git: Fetch" command
  • Confirm that the quick pick contains all existing remotes