Pathological performance on long call chain with list literal arguments

[munificent]

This (very large) statement hits the max number of tries in line breaking and fails to format ideally:

    selection.callMethod('enter').callMethod('append', ["svg"]).callMethod(
            'attr', ["class", "box"]).callMethod('attr', ["width", width])
        .callMethod('attr', ["height", height]).callMethod('append', ["g"])
        .callMethod(
            'attr', ["transform", 'translate(${margin[3]},${margin[0]})'])
        .callMethod('append', ["text"]).callMethod('attr', ["class", "label"])
        .callMethod('text', [(_, i, __) => groups[i]]);

The problem is that those list literals break each .callMethod() into a separate rule, so we can't easily split this before every . as one rule.

[munificent]

The forthcoming tall style doesn't have any performance problems on this code and produces:

    selection
        .callMethod('enter')
        .callMethod('append', ["svg"])
        .callMethod('attr', ["class", "box"])
        .callMethod('attr', ["width", width])
        .callMethod('attr', ["height", height])
        .callMethod('append', ["g"])
        .callMethod('attr', [
          "transform",
          'translate(${margin[3]},${margin[0]})',
        ])
        .callMethod('append', ["text"])
        .callMethod('attr', ["class", "label"])
        .callMethod('text', [(_, i, __) => groups[i]]);