kernel(precedence-ef): fix precedence_ef.mli's stale RG33/RG26/RG111(b) contract

The .mli's disposition doc still said a Class1-or-Class2 vigil is Omit
(RG 33) -- the pre-Task-16 mistranscription the register corrects at its
RG 33 entry. The .ml has read Class2 || Class3 since commit 079e332; the
.mli never caught up. It also named only vigil_suffix, when the vigil
test the code actually runs also checks vigil_prefix (the shape the
sanctoral bootstrap's four real vigils use); omitted RG 26's Class4
temporal-ferias-are-never-commemorated branch entirely, then asserted the
listed cases 'exhaust every representable shape', which was false as
written; and omitted RG 111(b)'s 'de festo II classis' rank floor from
admit's doc, describing only the privilege-override half of that rule.

Fixed all four: the vigil-omission bullet now names Class2/Class3 and
both slug conventions; the missing RG 26 branch is now documented,
folding the exhaustiveness claim into an accurate five-branch count;
admit's Sunday case now states the rank floor as a second, independent
condition.

The same stale RG 109(e) reading this interface carried ('leaves the
Advent and Lent Ember sets ordinary') also survived as the stated reason
september_ember_prefix exists, in the .ml, about 330 lines from the
already-corrected privilege_of comment it contradicts. Fixed there too:
September is broken out because it sits outside RG 109(e)'s three named
seasons entirely, not because Advent/Lent Ember needs excluding from (e)
-- it does not, per privilege_of's own corrected comment.

No behaviour change: both fixes are doc-comment-only edits to already-
correct code (precedence_ef.ml's disposition/admit implementations were
fixed in an earlier commit; only the .mli's prose and one earlier .ml
comment lagged).

1 files changed, 12 insertions, 4 deletions

diff --git a/lib/rites/rite_ef/precedence_ef.ml b/lib/rites/rite_ef/precedence_ef.ml
index 88f1a54..bfd3b93 100644
--- a/lib/rites/rite_ef/precedence_ef.ml
+++ b/lib/rites/rite_ef/precedence_ef.ml
@@ -87,10 +87,18 @@ let is_vigil slug =
 
    [september_ember_prefix] is broken out as its own name (rather than an
    anonymous list literal) because Task 9's [privilege_of] needs to test the
-   September set alone, RG 109 privileging it while leaving the Advent and
-   Lent sets ordinary (register lines 375-376) -- building [ember_prefixes]
-   from it rather than duplicating the literal keeps the two from silently
-   drifting apart. *)
+   September set alone: RG 109(e)'s three named seasons (Advent, Lent,
+   Passiontide, §4 "Commemorations") never include September, which sits
+   entirely in time after Pentecost under any reading -- so September Ember
+   days need their own separate privilege category, (d), regardless of how
+   (e) itself is read. CORRECTED (fix round 1, F1/F2): this comment
+   previously justified the split the other way round, claiming RG 109(e)
+   privileges September specifically "while leaving the Advent and Lent
+   sets ordinary" -- WRONG; see [privilege_of]'s own (e) comment below for
+   the full argument. The Advent and Lent Ember sets ARE privileged under
+   (e), the same as any other Advent/Lent feria; building [ember_prefixes]
+   from this constant rather than duplicating the literal keeps the two
+   from silently drifting apart. *)
 let advent_ember_prefix = "ef-advent-ember-"
 let lent_ember_prefix = "ef-lent-ember-"
 let september_ember_prefix = "ef-september-ember-"