Proposal: Read/calculate and store uncompressed size of compressed datasets

[natefoo]

A lot of TPV rules are based off input size, but input size is often a lie due to compression. This is likely a significant cause of over or underallocation of memory for jobs. I opened an issue over on the TPV repo but suspect this should probably be done internally in Galaxy especially if we don't want TPV to be expecting data to exist on disk.

As noted in the issue, the uncompressed size of gzipped data can be stored in the last 4 bytes of the file, but it is sometimes empty, in which case the only way to get it is to decompress.