Document other weight meta-data fields

We currently document the number of parameters in the Weights meta-data and we display these on our summary tables. Other values might be of interest for users:

• FLOPs. @Chillee 's FLOP counter might be useful for this -- done in https://github.com/pytorch/vision/pull/6936 and https://github.com/pytorch/vision/pull/7074
• Model size in MB. This will usually be ~= 4 * num_params, except for quantized models. Either-way, it's worth computing and documenting directly (Note: https://github.com/pytorch/vision/pull/6936 added docs for the file size, which may (or may not?) be different)
• Memory requirements for 1 image (forward and backward pass). This can be estimated manually but can quickly become untractable, so it's best to do that automatically with a script. We should be able to use memory_stats module for this (accounting for the caching allocator, etc). I'll need to think about this more to make sure what we report is meaningful.

CC @datumbox

Start by tracing the existing Weight metadata and the summary-table generation, then compare the file-size documentation added in pull requests 6936 and 7074. Define how model size should be measured, especially for quantized models, and investigate torch.cuda.memory_stats for forward and backward memory estimates. Done means the selected measurements are computed consistently and documented for users.