Fixes the MSize comparison when negociating msize with server

[simonferquel]

The msize negociation was wrong.
According to the man page http://plan9.bell-labs.com/magic/man2html/5/version, and Linux kernel 9pfs implementation http://git.kernel.org/cgit/linux/kernel/git/ericvh/v9fs.git/tree/net/9p/client.c#n980, the flow of version is:

• client send its "HARD" max size
• server compare it to its own "HARD" max size and respond with the minimum of the 2 values
  -> in the response the msize cannot be greater than the original client max size (if server has a greater value, it scales down to full fill the client constraint, if its lower, it tells to the client that it can't accept such a large msize value)
• client and server agrees to scale down to the lowest "Hard" max size

[codecov-io]

Current coverage is 15.95% (diff: 0.00%)

Merging #28 into master will not change coverage

@@             master        #28   diff @@
==========================================
  Files            13         13          
  Lines          1128       1128          
  Methods           0          0          
  Messages          0          0          
  Branches          0          0          
==========================================
  Hits            180        180          
  Misses          900        900          
  Partials         48         48          

Powered by Codecov. Last update 5a23906...b8387ce

[stevvooe]

LGTM