Legacy line-break placeholders still keep FEFF until IME proof exists

What happened

After proving that the v2 DOM bridge already tolerates -style line-break placeholders, the tempting next move was to drop FEFF from legacy isLineBreak rendering too.

That was too broad.

The evidence only proved:

• renderer shape
• basic focus / selection sanity

It did not prove IME composition safety on the empty-block start-of-block path.

What fixed it

The honest legacy policy stays conservative:

• line-break placeholders still render
• FEFF is still retained on the empty-block start-of-block line-break path
• non-linebreak zero-width cases also keep FEFF

That keeps legacy behavior aligned with the known IME-risk comments while still making the future policy line explicit.

What this means

The larger split still stands, but only as a future v2 renderer policy:

• the bridge already supports line-break placeholders
• the remaining unknown is IME behavior on that path

So the next proof is not “can the bridge handle ?”

It is:

• can empty-block IME composition work without FEFF on the line-break path?

Reusable rule

Do not widen a zero-width rendering policy beyond the evidence.

If the proof only covers DOM shape and selection sanity, keep the IME-sensitive path conservative until IME behavior is proved directly.