Intersection Over Union (IoU) should support Shape N x C_1 x C_2 x ... x C_n

Nobody has claimed this yet.

## 🐛 Bug

I was writing a Pull Request for Sorted AP metric and I figured "let's use torchmetrics's IOU function!". Unfortunately, unlike the [JaccardIndex](https://lightning.ai/docs/torchmetrics/stable/classification/jaccard_index.html), torchmetric's [IntersectionOverUnion](https://lightning.ai/docs/torchmetrics/stable/detection/intersection_over_union.html) does not take multiple dimensions.

Infact, it only allows the very limiting shape of (N, 4). This is not an oversight; it's a major flaw hurting torchmetric's credibility because IoU is the most well-known computer vision metric. A user will first look for IntersectionOverUnion and NOT JaccardIndex and chances are that they'll be disappointed.

I propose that IntersectionOverUnion be a wrapper over [MultilabelJaccardIndex](https://lightning.ai/docs/torchmetrics/stable/classification/jaccard_index.html#torchmetrics.classification.MultilabelJaccardIndex)

I am willing to make a PR for it.

### To Reproduce

See any example from here. [IntersectionOverUnion](https://lightning.ai/docs/torchmetrics/stable/detection/intersection_over_union.html)

Start with the IntersectionOverUnion entry point and compare its documented input shape with the MultilabelJaccardIndex behavior linked in the issue. Verify the intended N x C_1 x C_2 x ... x C_n support against the existing IntersectionOverUnion examples, then add coverage showing the expanded dimensions work without breaking current usage.

Written by the indexing model from the issue text.