ENH: add gzip/bz2 compression to relevant read_* methods

[gfairchild]

This issue is a branch off of #11666, which implemented compression support for read_pickle. There are still a few other read_* methods that could possibly benefit from compression support. Looking at the I/O API reference, this jump out at me:

• read_json - This can definitely benefit from compression. I've stored very large gzipped JSON files before. As a general rule, any read_* method that supports any kind of plaintext format should support compression.
• read_stata- I don't use Stata, but it looks like a .dta file is not a plaintext file. Is it naturally compressed, or can they be compressed significantly like pickles?
• read_sas - I've also never used SAS, and like Stata's .dta files, it looks like .xpt and .sas7bdat files are both some binary format. Can they be compressed well?

[jreback]

most important is json

[bashtage]

Stata is not compressed but is just a fairly plan binary file format. This said, I don' t think there is much of a reason to add compression methods since the output file wouldn't be usable in Stata (presumable the reason to output in this format) without manual decompression.

[jreback]

IIRC read_sas has internal compression as well? or is it a different file extension?

[gfairchild]

In this case, it looks like read_json may be the only method that needs compression support added.

[jreback]

@gfairchild want to take a stab at this? should be fairly straightforward as you can pretty much reuse the infrastructure (mainly just passing the compression arg thru). This is really just a couple of tests as well.

[gfairchild]

I'd be happy to. Just got to find the time. Maybe I can do it this weekend.