Optimize doResize method with binary search

[JustSNguyen]

This pull request addresses issues discussed in #6112 . Currently, if an uploaded image exceeds a certain size threshold, BlueSky reduces the image quality by 10% and checks if the new size falls within the threshold. If not, BlueSky repeats this step until the image quality reaches 0. There are two issues with this approach:

1. As mentioned in Proposal: Be smarter about quality level when converting images to JPGs #6112, even a 1-2% decrease in quality can significantly reduce the image size, so the 10% decrement is somewhat overkill.
2. The current solution could be optimized, especially for larger images. For these, the current approach requires multiple iterations to reach the target size, which is not ideal.

My proposed solution is to use binary search to find the maximum quality level that keeps the image size within the limit. This approach will prevent too much loss in image quality. Not only that, I also ran some benchmarks, and it appears that the current solution may perform slightly better for smaller images. However, for larger images, the binary search method does much better. That said, I only conducted a few tests here, so additional testing may be needed to confirm the difference.

Image Size (KB)	Current Method (ms)	Binary Search Method (ms)
1148	844	1040
1911	5454	6477
3011	5530	3396
3230	6079	5402
10609	12210	4141
14765	26390	10213

[gaearon]

This makes sense. Let's do it on all platforms?

[JustSNguyen]

I have updated the PR to also support this for the mobile version @gaearon.

[auroursa]

Excellent! Uploading photos taken with an iPhone（obviously compression is needed） clearly shows that the new algorithm improves image quality noticeably, with a shorter processing time as well.

[gaearon]

This is changing lib/media/manip but I'm seeing some code go through a similar path in src/state/gallery. Would you mind checking how it works today? It seems like maybe it's forked between the image picker vs inline camera, but I'm not sure.