Fix unstable own line comment in a for-in or for-of head - #19895

Description

js/for-of/comments.js is one of the unstableTests entries in tests/config/format-test/failed-format-tests.js. This makes it stable and removes it from the list.

An own line comment in a for-in or for-of head is unstable because printing it turns it into an end of line comment, and the two placements take different branches:

// Input
for (x
//c
in y);

// Prettier stable (first format)
for (x in //c
y);

// Prettier stable (second format)
for (x in y); //c

The comment starts out on its own line between left and right, so it becomes a leading comment of right and prints right after in. That puts it at the end of a line, so the next format makes it a trailing comment of left instead and moves it out of the head.

There is no line inside the head to keep it on, so it ends up outside either way. This attaches it to left immediately, which is where the second format already put it. Block comments were unstable the same way and are fixed by the same change:

for (x
/*c*/
in y);   // was: for (x in /*c*/\ny);  then: for (x /*c*/ in y);

Note the old first format was not just unstable but mis-indented — y); was printed at column 0.

The handler is registered in the own line list only. End of line and remaining comments already attach to left by default and are unchanged:

for (x in //c
y);                  // -> for (x in y); //c   (unchanged)
for (x /*c*/ in y);  // -> for (x /*c*/ in y); (unchanged)
for (x in /*c*/ y);  // -> for (x in /*c*/ y); (unchanged)

ForStatement is not included — its head already breaks, so own line comments there are stable and print in place. Comments before left, and comments before the body, are also untouched, the latter being what the existing handleForXStatementComments already handles.

The resulting fixpoint keeps the comment on its own line, so it differs from the old second format output for these inputs. Since the old first and second formats disagreed, neither was a settled output.

One related case is still unstable, unchanged by this PR and failing the same way on main:

for (
/*c*/
x of y);

It has no preceding node, so it needs different handling than this and I would rather not fold it in here.

Across the whole tests/format run this changes exactly one snapshot, the fixture itself, and the failing set is the same before and after.

Checklist

• I’ve added tests to confirm my change works.
• (If changing the API or CLI) I’ve documented the changes I’ve made (in the docs/ directory).
• (If the change is user-facing) I’ve added my changes to changelog_unreleased/*/XXXX.md file following changelog_unreleased/TEMPLATE.md.
• I’ve read the contributing guidelines.
• I did not use AI to generate this PR.
• (If the above is not checked) I have reviewed the AI-generated content before submitting.