move read_t to binary.py namespace

[bolliger32]

Addresses #744 by moving read_t such that it is in the module namespace. I'm not aware of why read_t was imported within the read function to begin with, so I'm not sure if this is going to create other issues. Someone with more of an understanding of pyclaw should probably take a look at this before anything gets merged.

[coveralls]

coverage: 46.613% (+0.02%) from 46.591%
when pulling 7a527f6 on bolliger32:bugfix/read-t
into 60de625 on clawpack:master.

[ketch]

Yes, it seems that several bits of code expect read_t to be available in the way that this PR would make it. Indeed, prior to 6e284b9, it was available there. I think when that change was made, we must not have realized that code outside of binary.py depended on this.

Thanks for catching this issue and fixing it!