Also fix formatting a bit.

It should return the file descriptor instead of file content as string.

This implementation does not consume RAM Memory, so it is safer to
download big files.

When you call file_descriptor.close, it will remove the file from tmp
folder automatically.

There is no need to read and store downloaded file. Returning the location is
good enough.

Rest of file requires hexdigested file.

There is no need to read whole file into memory on library level. So calculate
sum in chunks and give request method file descriptor.

Indeed, some day documentation shall be written, as soon as API will be defined.

HTTPConnection accept port as none, there is no need to play with it.

The select method  must return the file from shacache.
And the download method should return the file content instead of path.

If the directory has more than one entry, the lib must raise such Error.

httplib does not handle the download as expected.
The implementation of urlretrieve is much better and consistent, it does
the download
writing in chunks instead of loading all the file content into memory.

Also, the httplib can easily break the file content, if the client does
not
handle it. And this lib must be very simple.

The timeout is not in favor of httplib GET method.

Based in the above reasons, httplib was replaced to
urlretrieve, once there is no reason to reimplement what already exists.

Because the user can just use the 'select' method to retrieve the information
from shadir server.

Otherwise, all the connections would still alive and this is not a good
behavior.

Once we need to close the connections it is not possible to return the
content of getresponse method, the method read would return an empty
string instead of the correct data.

And also the data returned by shadir is a json content,
to reach a better usage of libnetworkcache the data returned must be
loaded by json library, in other words, it must be a list.