Tests: negative assertions that could not fire, and a test that pins inbound_count

[muellerei]

Three assertions checked that a message is absent, and the message no longer exists, so they could never fail:

• test_journal_block_reports_its_position: "Could not find or create" went with the top-level fallback in 8926e36. The line is redundant: json.loads(result.stderr) above it already requires stderr to be one JSON object, so any extra warning fails the test first. Removed.
• test_keep_block_ids and test_keep_ids_every_writer: "cannot preserve" went with the note in 5860462, when top-level blocks began to keep their ids. They now check that the note which does exist, the dropped-id:: one, does not appear; their sibling tests assert the same text is there when ids are dropped. stderr is not empty in every case (add-journal-block --file prints an unrelated note), so a blanket emptiness check would break on that.

test_get_page_stats_does_not_report_zero_inbound_links_after_an_error looked for "inbound_count" in stdout, and no test expected the field anywhere: renaming it left the whole suite green. A new test reads the stats when the scan can read every page and expects the field and its value.

Checked

• 3266 tests, ruff check ..
• Printing the dropped-id note despite --keep-ids passes the old assertions and fails the new ones, in all eight cases.
• Renaming inbound_count passes the whole suite as it stands on main and fails the new test.
• Printing the old fallback warning fails the journal test with or without the removed line.

Not changed: (empty), Added block-ref, Done. and one note in init_config have the same kind of loose wording check; each has part of its claim covered elsewhere, and none has let anything through so far.