xterm.js backend: revisit addon-unicode-graphemes and addon-clipboard (OSC 52)

Tracking the two `@xterm` addons the xterm.js backend deliberately does not use yet, moved out of `crates/tui-test/assets/xtermjs/README.md` so the vendoring notes stay short.

### `@xterm/addon-unicode-graphemes`

Would replace the pinned `@xterm/addon-unicode11`. Not adopted because it:

- is marked experimental by its own package description,
- needs an `atob` the QuickJS host does not provide, and
- measures two cases differently from alacritty:

| Input | alacritty | v11 (pinned) | v15-graphemes |
| ----- | --------- | ------------ | ------------- |
| `👨‍👩X` | 5 | 5 | 3 |
| `👍🏽X` | 5 | 5 | 3 |

Agreeing with alacritty on width is why unicode11 is pinned where it is: a disagreement puts every cell after an emoji in a different column on the two backends, which moves what `cells`, the locator, and the SVG renderer report.

**Revisit when** it is no longer experimental, and adopt only if it still agrees with alacritty on that table.

### `@xterm/addon-clipboard`

Implements `OSC 52`, which tui-test does not support on any backend today — alacritty parses it and raises `ClipboardStore`/`ClipboardLoad`, and the listener in `terminal/alacritty.rs` drops both.

Adding it to xterm.js alone would make things worse rather than better: `OSC 52;c;?` is answered by nothing today, and this would make one backend answer it, so the same sequence would behave differently depending on the selected backend.

**Order matters here:** `OSC 52` should land in the `Emulator` contract and the conformance suite first; this addon is then how xterm.js implements its half.

Start with the vendoring notes in crates/tui-test/assets/xtermjs/README.md and the OSC 52 handling in terminal/alacritty.rs. Read the Emulator contract and conformance suite before considering addon-clipboard; verify that OSC 52 behaves consistently across backends. Revisit addon-unicode-graphemes only when it is no longer experimental and still matches alacritty's widths in the listed cases.

Written by the indexing model from the issue text.