#1441: Accept can-prefixed positive tests

[zaguzovmaksim0-hue]

Closes #1441.

Aligns unit-test-is-not-verb with the sibling bad-test-name rule. Positive test names beginning with can- are now accepted explicitly; all other positive names continue through the existing OpenNLP VBZ verb check.

This keeps accepts-* and ordinary singular-verb names working as before while removing the contradiction where bad-test-name requires/permits can-* but OpenNLP tags can as a modal (MD) and the Java lint rejects it.

The existing positive-name parameterized test now includes can-read-bytes as a regression case. git diff --check is clean.

The commit was created through GitHub and is cryptographically marked Verified.

[yegor256]

Thanks — this does resolve the collision #1441 describes, and it is the direction the issue proposes. One concern before merging: the can- prefix is now a string literal inside LtTestNotVerb.isVerb, while the set of allowed prefixes is really owned by bad-test-name.xsl (^(can|accepts)-) and by Vocabulary. Two places now have to be kept in step, which is how the two lints drifted apart in the first place. Teaching Vocabulary that a modal followed by a verb is acceptable, or reading the prefix set from one shared place, would keep them from diverging again.

Assigning to @yegor256 to decide which of the two resolutions the issue offers he wants.

---

Generated by Claude Code