[README] Guide a fresh clone through its first build - #745

Get a new contributor from clone to running app without guesswork. A fresh clone needs a one-time setup step (make bootstrap) that was not documented anywhere, and skipping it, or having it cut off, failed with errors that pointed elsewhere, such as a missing script, rather than at the step that was missed.

Changes

• Document the setup step: the README says to run it before the first build, in Xcode or on the command line, and gives the shortest path from a fresh clone to a running app. This replaces the 2023 build script from discussion I can't get the built app to run: GitX quit unexpectedly. #366 the README linked to, which the Makefile now covers.
• Make clear that signing is optional: Debug builds, from Xcode or make run, sign ad-hoc and need no setup, while the README read as if every build needed a Dev.xcconfig. make Dev.xcconfig is now the one way to set it up, since it reads everything off the certificate and says what is missing; writing the file by hand and digging the team ID out of Keychain Access are gone.
• Name the setup step when it was skipped: make deps and the warning before each build now say to run make bootstrap, instead of failing on a missing file or suggesting a target that leaves the dependencies unbuilt.
• Let make bootstrap finish a run that was cut off: an interrupted submodule update leaves a clone without its first commit, which every later update fails on, and clones that were never checked out, which every later update skips. scripts/repair-submodules.sh removes the first for the update to clone again and checks out the second; neither holds anything to lose, and a submodule in any other state is left alone. The check before each build reports both.

Test plan

• In a test clone, recreated both states as an interrupted run leaves them (a cut-off openssl under objective-git, and git-diff-image and screenShotScript never checked out): the old update failed with "Unable to find current revision" and the old check was silent; the new check lists all three (and fails on CI), and one make git-submodule-sync repairs them, leaving a clean tree.
• The script leaves alone a local edit, an unborn HEAD that still has refs, an emptied index beside an untracked file, and deleted files with the index intact; its check is silent on healthy checkouts.
• In a fresh worktree with no submodules checked out, make deps stops and suggests make bootstrap. CI's own make deps steps still pass, on a checkout that has its submodules.