Use a different unit rather than 'level' for pathology image resolution

Is your feature request related to a problem? Please describe.
Some of the WSI-related Dataset features use 'level' as the unit of down-sampling for patches. This is potentially problematic because level is not an absolute or relative value to anything in particular. 1 image might have 2 levels with level 0 representing full resolution and level 1 representing a 2.0 downsample factor. A different image might have 4 levels with downsample factors of, say, 1.0, 4.0, 8.0, 32.0. This means that specifying a 'level' in a Dataset that contains images from different sources could, at worst, cause errors, because the level specified is not present in all of the images or cause bugs since the levels in two images corresponds to very different resolutions.

Describe the solution you'd like
In an ideal world, we'd use magnification or pixel size (in physical units), but may not always be feasible because this information is not always in the metadata. Downsample factor might be a better solution. The API can use the metadata to find the level that matches the required factor, but perhaps we need to allow the user to choose the units ('pixels per mm', 'magnification', 'downsample factor', 'level') and also provide a fall-back to cater for the lack of the required unit information. e.g. If they specify downsample as 4x and the image only has 2x or 8x levels then it could either downsample down to 4x from 2x automatically or upsample from 8x to 4x.

No files, tests, or entry points are named. Start by locating the WSI-related Dataset features and the metadata used to select image levels, then compare how level and downsample factor are represented. The issue does not select a final unit or fallback behavior, so completion would require an agreed API and matching tests.