Flips the second amendment on the text-first record - the 2026-09-04
sui-gcm one, that a picklist is a rendering of the expression string
and the component owns no second representation - from proposed to
accepted. The already-accepted sui-qay amendment above it and the
record's own top-level status are untouched.

Following the sui-qay flip's precedent in this same file, the change
moves the header amendment status line, the section heading, and the
section's proposed/accepted marker. It also applies the wording that
sui-gcm's review asked the flip to carry: the sentence about holding
no second copy of the grammar is scoped to the grammar's spellings,
which is what the code guarantees now that eligibility is the
grammar's too.

Re-checked against the tree before flipping, on predicator 9.2.0.
simple/2 still answers three ways and keeps them apart; every
spelling still round-trips through to_source/1; the degrade is still
get_env plus ensure_loaded? plus function_exported?, still answering
:outside; the picklist mode, the fallback order and both switch
affordances are real in expression_input.ex; and the eligibility
table the amendment named as its one exception is gone, exactly as
the amendment said it would be.

Accepting the amendment elevates the sui-94o note inside it, whose
standing rides with it, so its per-kind table was re-verified against
live Predicator.Simple.operators/1 rather than taken on the earlier
review's word. Every row's "what changed" holds against the deleted
local table, the display-order claim holds, and the vocabulary still
names seven kinds with no relative_date among them. The note's
preamble is re-tensed here, because it asserted the amendment was
still proposed.

Checked against 8119f19 and 950c070, the two commits that rewrote
docs/wire-format.md after several records merged. This section cites
that document nowhere and neither commit touches the expression
modules, so no claim in it is affected.

Refs: sui-2hy