Figure out a better way to display file sizes

[Munter]

The current crazy bloated file size reporting is caused by the fact that you can never write more data to the read buffer than you originally said the file would contain in the stat operation.

During the stat operation the file might not be in cache, thus not have a valid file size.

The current hack I'm using is to report 1mb + 10 * file size to be sort of certain that a transpilation output won't be bigger than the stat operation said. Then I simply return an empty read callback while incrementally reading, which correctly stops the read operation from reading all the wrongly reported non-existent bytes.

It would be much better to have a way to simply show no file size and let the unknowns be unknowns. The problem is I haven't figured out a way to do this yet while still allowing file reads that are correct.

[Munter]

https://github.com/assetgraph/skrin