lib_git.sh has no tests for failure paths: network timeout, non-fast-forward, dirty worktree

[codeforester]

Problem

lib/bash/git/tests/lib_git.bats tests the happy path of git_update_repo but does not cover:

1. Network timeout / remote unreachable — git fetch hangs or fails; git_update_repo should return non-zero and not mutate the worktree.
2. Non-fast-forward update — the remote has diverged (force-pushed or rebased); --ff-only will fail. The test should verify the local branch is left unchanged.
3. Dirty worktree — basectl update explicitly checks for a clean worktree, but this is checked in update.sh before calling git_update_repo. There is no test that validates the integration between the clean-worktree guard and the git library.
4. Missing remote — no origin remote configured; git_update_repo should fail clearly.

These are precisely the scenarios that matter in real CI and offline development environments.

Proposed Tests

@test "git_update_repo fails clearly when remote is unreachable" {
    # set remote to a non-existent URL
    ...
    run ... git_update_repo "$repo" "" "$branch"
    [ "$status" -ne 0 ]
    [[ "$output" == *"fetch"* || "$status" -ne 0 ]]
    # local branch should be unchanged
}

@test "git_update_repo fails on non-fast-forward and leaves branch unchanged" {
    # set up diverged history with forced push simulation
    ...
}

Why This Matters

basectl update is a command users will run when something is already wrong (they need a bug fix). A failure in the update path — especially a silent one that partially mutates the repo — is the worst time for the tool to be unreliable.