(RFE) Add size and other metadata in LATENT dictionary.

### Feature Idea

It would be nice to add some metadata about a LATENT sample to the dictionary for this type. This could include things like the pixelspace extent, name of the VAE class that encoded it, metadata from the VAE alggorithm itself, etc.

In my particular use case, I want to resize an image to fit the pixelspace for a particular LATENT samples. Currently the only way to do this is to actually decode the latent into an IMAGE, which is inefficient. The sample size and other constants necessary to calculate the correct size are hard-coded and not available as VAE metadata at runtime, it seems - and even if they were, there would be no way to know which VAE encoded the sample.

### Existing Solutions

No solutions are currently available.

### Other

_No response_