Make rewrite source-note fetch best-effort so shifts never abort wholesale

[a-churchill]

When a rewrite (rebase/cherry-pick/squash) had a source commit with no authorship note anywhere and the refs/notes/ai fetch from a remote failed (off-VPN, broken SSH in the daemon environment, unreachable remote), fetch_missing_notes_for_commits returned Err and the ? at the rewrite call sites aborted the entire shift — commits whose notes were sitting in the local notes-db lost them too. Under the HTTP notes backend this seeded a cascade: each rebase's noteless outputs became the next rebase's noteless sources.

Route the four shift call sites through a best-effort wrapper that logs the fetch failure and proceeds, so locally available notes always migrate. Sources that stay missing simply carry no note, same as before, but the loss no longer widens to the whole rewrite.

Regression tests run a real rebase under an HTTP-backend daemon with a noteless plumbing-created source commit: one with an unreachable remote (previously aborted with "failed to fetch authorship notes for source commits [...]: exit code 128"), one control with no remote.

Note this is meant to be complimentary to #1975

---

[CLAassistant]

Thank you for your submission! We really appreciate it. Like many open source projects, we ask that you sign our Contributor License Agreement before we can accept your contribution.
_{You have signed the CLA already but the status is still pending? Let us recheck it.}

[devin-ai-integration]

✅ Devin Review: No Issues Found

Devin Review analyzed this PR and found no bugs or issues to report.