Labelled align* takes an equation number (starred envs should be unnumbered regardless of identifier)

[mmcky]

Found while adopting qe-v9 in the LaTeX→MyST converter (QuantEcon/claude-latex-to-myst#186). Verified against myst v1.10.1 (qe-v9).

Summary

A labelled align* takes a real equation number. align* is unnumbered in amsmath, and #81's own description says "align* and friends: unnumbered" — but that only holds when the node has no identifier.

Reproduce

With numbering: {book: {enabled: true}}:

$$
z_0 = 0
$$ (eq-base0)

(eq-star-outer)=

$$
\begin{align*}
a &= b \\
c &= d
\end{align*}
$$

$$
z_1 = 1
$$ (eq-base1)

Built AST:

node	enumerator
eq-base0	1
eq-star-outer (the align*)	2
eq-base1	3

The starred environment consumed number 2 and pushed everything after it along. Without the (eq-star-outer)= anchor the same block is correctly unnumbered.

Cause

packages/myst-transforms/src/math.ts sets the unnumbered flag only when the node has no identifier:

if (node.enumerated == null && !node.identifier) node.enumerated = false;

Once a label is attached the !node.identifier guard fails, so the starred env falls through to normal numbering and also loses node.rows.

Expected

A starred environment is unnumbered regardless of whether it carries a label. In LaTeX, \label inside align* is legal — it just refers to whatever the enclosing counter last was — and it certainly does not cause the environment to be numbered.

Why it matters downstream

This is a live shape: book-dp1's ch_ctime.tex has a labelled align* (eq:vgctp), the only one across the three QuantEcon books. The converter currently emits starred envs as a {math} directive with :enumerated: false to force the correct behaviour; that workaround exists precisely because of this. If this is fixed we can drop it and pass starred envs through like the non-starred ones.