fix(of): print the OF's own transcribed Latin/English names, not EF's

day_line_of/readings_line_of resolved every displayed name through
lang/la.ini -- data documented as "transcribed from the 1962 Missale
Romanum", i.e. EF vocabulary, keyed on slug. 18 of 222 shipped OF
sanctoral slugs collide with an EF slug there and all 18 printed the
WRONG (1962) title (e.g. "S. Marthae Virg.", a rank word the 1969
reform abolished, silently overriding amendment 011's own "Ss.
Marthae, Mariae et Lazari"); the other 204 had no entry and degraded
to a bare slug, so none of the 208 titles calendar-2002.sexp itself
transcribes (cited against the 2002 Missal) was ever reachable.

Add observed_name_of: prefer the observed celebration's own
Celebration.names (looked up by the requested language code), fall
back to the lang/*.ini table only on a miss. Under --raw this is a
no-op (Lang.raw's own code, "raw", is not a valid 2-letter ISO code,
so the names lookup always misses and falls through to the identity
table exactly as before). EF's day_line/readings_line are untouched --
EF's Celebration.names is almost always empty by design, lang/la.ini
is its real, independently-cited name table.

Regression net, two layers: test_lang_coverage.ml gains a test walking
the real assembled OF layer's entries, asserting that every genuine
lang/la.ini collision resolves to the OF's own name, never the EF
table's; test/cli.t pins one real collision (Martha, 29 July) through
the actual CLI in both la and en.

1 files changed, 12 insertions, 0 deletions

diff --git a/test/cli.t b/test/cli.t
index 8671694..86291a6 100644
--- a/test/cli.t
+++ b/test/cli.t
@@ -232,6 +232,18 @@ The OF rite resolves too, same two commands, its own vocabulary:
   $ colitur readings --rite of 2026 | head -1
   2026-01-01 of-mary-mother-of-god | Num 6:22-27 | Luc 2:16-21
 
+Fix wave I1 (final-review.md): a sanctoral-origin OF day names through its
+OWN transcribed title (calendar-2002.sexp, amended by 011's own rename),
+never through lang/la.ini's EF (1962) table -- 29 July, Martha, is one of
+the 18 real slug collisions between the two vocabularies the review found;
+la.ini's own entry for "martha" is the pre-2021 EF title "S. Marthae Virg.",
+and printing it here would be exactly the regression this guards.
+
+  $ colitur day --rite of 2026 | grep '^2026-07-29 '
+  2026-07-29 wednesday ordinary-time 17 martha memoria-obligatoria white Ss. Marthae, Mariae et Lazari
+  $ colitur day --rite of 2026 --lang en | grep '^2026-07-29 '
+  2026-07-29 wednesday ordinary-time 17 martha memoria-obligatoria white Saints Martha, Mary and Lazarus
+
 An unrecognised --rite value is a usage error, not a silent fallback to ef
 (exit 2):