I think this can be made to be an example, not sure it's worth the maintenance overhead. But I might be too strict... Your call...

IMHO, it's quite a trivial block of code that

• shouldn't require much maintenance, as its dependencies are quite limited to core classes and methods that shouldn't change in time
• easy to implement in trunk
• it's compatible with any type of geoJSON geometries, even when mixed in the same file or geometry collections
• enables the classical feature zoom to layer for geoJSON sources (the only source which bounds might be known without messing with the arbitrary metadata)

From my point of view, the benefits of adding this method are way above the costs 😄

NOTE: I've updated the snippet I made yesterday in a 5-minute break so that it can avoid extra, unneeded, steps

Faster (and ES6+) version

async function _getbounds() {
	const bounds = new maplibregl.LngLatBounds();
	const data = await this.getData();
	const coordinates = data.features
		.map(f => f.geometry.coordinates.flat(Infinity))
		.reduce((acc,cur) => [...acc, ...cur],[]);
	for(let i = 0; i < coordinates.length - 1; i += 2){
		bounds.extend([coordinates[i], coordinates[i+1]]);
	}
	return bounds;
}

maplibregl.GeoJSONSource.prototype.getBounds = _getbounds;

As I said, feel free to open a PR, I don't have strong feelings either way...

Done: #5403

🤓