CLN: consolidate compression inference as much as possible

[jreback]

we have 3 copies of the compression inference routines in parser.py, parser.py, and io/common.py. See how much we can reduce this (some is necessary as sometimes we need to actually return a file handle or need to read in and re-wrap).

xref #12175

[jreback]

cc @lababidi

[jreback]

@gfyoung if you are looking for something to do :>

I think we can blow away _wrap_compresed in parsers/io/parsers.py and incorporate any such functionaility in io/common.py/_get_handle

there is also compression handling in parser.pyx but I think this might be harder to remove (in favor of _get_handle, though maybe if we pass an argument might work).

xref #13317

[gfyoung]

Haha. Thanks @jreback for the mention. I'll take a closer look when I have the time. 😄

[lababidi]

@jreback I got the following error when attempting to test:

TypeError: assert_almost_equal() got an unexpected keyword argument 'check_dtype'```