fix: no-DOI entries were judged against the wrong paper (0.6.1)

[musharna]

0.6.0 accepted CrossRef's first search hit for any entry without a DOI and compared the byline against it. On real reference lists most of those hits are a different paper.

document	entries	0.6.0	this branch
Thorne 1998 (Syst. Biol.)	26	6 A, 7 U, 1 C	8 U
Sanderson 2002	47	12 A, 12 U	1 B, 1 C, 11 U
Parham 2012	146	41 A, 5 B, 36 U	3 B, 1 C, 24 U
Word draft (9 DOIs)	16	3 A, 3 U	2 A (both DOI-path), 6 U

All 59 Tier A findings on the three published lists were false. The residual U rows are books, software manuals and the editor byline; the B rows are one-year online/print differences.

• query with the whole printed reference when no title was parsed
• search_hit_is_cited_work: title evidence + cited author somewhere in the byline (+ year when the verdict would be author order); otherwise Tier U naming the closest match
• a rejected hit no longer feeds the retraction lookup
• splitter: byline wrapped after an initial, publisher tail lines, running heads

Every new test was seen failing on the old code; the author-order Tier A test passes before and after. Bumps to 0.6.1 (incl. the Action pin); tag v0.6.1 after merge.