Update zero line logic for sploms

[alexcjohnson]

#2936 fixes zero line logic for regular cartesian plots, but large sploms do their own thing for performance. As @etpinard points out, this should be udpated too:

No need this do in this PR (opening a new issue would be fine), but we'll need to propagate these changes to large-sploms' regl-based grid line logic:

plotly.js/src/traces/splom/base_plot.js

Lines 171 to 183 in b5f8b23

	// just like in Axes.doTicks but without the loop over traces
	function showZeroLine(ax) {
	var rng = Lib.simpleMap(ax.range, ax.r2l);
	var p0 = ax.l2p(0);
	return (
	ax.zeroline &&
	ax._vals && ax._vals.length &&
	(rng[0] * rng[1] <= 0) &&
	(ax.type === 'linear' || ax.type === '-') &&
	((p0 > 1 && p0 < ax._length - 1) || !ax.showline)
	);
	}

where I didn't just reuse the axes.js logic because I wanted to avoid looping over the traces (to compute that hasBarsOrFill boolean). Thinking about this again, looping over traces isn't really a performance hit for sploms (I can't really imagine a user wanting to plot more 10 splom traces on the same graph).

So it would be worth trying to expose and ♻️ this logic here for large-sploms.