docs: record item 36, and mark 24/25 done in the table

Two 0.8.0 defects could not be given regression tests because
test_mainwindow has no notmuch database, so threadLoaded never fires:
the queued reply repainting a blanked pane, and a local sync reporting
itself as a background one. Both were verified by hand and by a
standalone model, and the first guard is still untested.

tests/notmuchfixture.h already does what is needed and test_notmuchworker
already uses it; test_mainwindow does not. So the item is wiring, not new
machinery.

Records that a corpus built from real mail was considered and rejected:
the rule against personal data in fixtures forbids it, hooks enforce it,
the repo is public, and one surviving In-Reply-To header names a real
correspondent. Minimal hand-written .eml files reproducing a message's
shape are the supported route, as truncated.eml already is.

Deferred by the user until a defect needs it. Wiring it with nothing to
test proves nothing.

The table also still showed items 24 and 25 open; both shipped in 0.8.0.

Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>

0 files changed, 0 insertions, 0 deletions