Speed up stale image checker

[Eric-Arellano]

#2528 improved the stale image checker to be "fast enough" to run occasionally, but the program is still quite slow due to starting up ripgrep 3450 times. It takes several minutes on my mac.

I suspect a faster approach would be to read the markdown to extract all used images, such as by using the link checker, then compare that to the images we do have on disk. It would avoid subprocess calls. However, it's possible this solution would overwhelm system memory.

[arnaucasau]

What do you think about integrating the stale image checker with our alt text image checker in https://github.com/Qiskit/documentation/blob/main/scripts/js/commands/checkImages.ts? I think we could use the unified plugin to find the stale images:

documentation/scripts/js/lib/markdownImages.ts

Lines 31 to 39 in be6687b

	visit(tree, "image", (node) => {
	// Sphinx uses the image path as alt text if it wasn't defined using the
	// :alt: option.
	const imageName = last(split(node.url, "/"));
	if (!node.alt || node.alt.endsWith(imageName!)) {
	imagesErrors.add(`The image '${node.url}' does not have alt text.`);
	}
	});
	visit(tree, "html", (node) => {

[Eric-Arellano]

That might be helpful to use the stale image checker. However, the stale image checker doesn't check historical API docs. Also, I realize we could probably implement #2538 such that we no longer even need to worry about checking API docs. If true, that would speed up this overall program enough automatically that we could avoid spending more dev resources on this ticket.

[Eric-Arellano]

Indeed, #2541 fixes the performance problem 🚀

It's blocked by #2540, which I want to fix because that is a total gotcha for future contributors.