sanity check length of recovery seed

[AdamISZ]

The issue that this addresses starts in #190 and continues/finishes in the reddit thread.

The fix is trivial, just checking that the seed phrase entered is 12 words. While doing it, I noticed that .split(' ') has the unfortunate effect of treating extra spaces as extra words (so 'a b c'.split(' ') comes out as ['a',b','','c'] which is undesirable if nothing else because it gives a misleading error). So I've edited it to .split() because this will split on any amount of whitespace, which is a more expected behaviour.

A perhaps more 'meta' conclusion from all this is to look again at sanity checking user input generally, and don't assume that used libraries (slowaes, old_mnemonic) are doing sanity checks at the lower level. I personally had never even looked at that code (even slowaes, although I used it before I never tried to use the high level 'encryptData' which does the padding for you).

[AdamISZ]

Oh, and I forgot to mention that the bug is more serious than just "you get a null seed if you press Enter".

If the word list passed to mn_decode is length 0 or 1 or 2, the function will always return '', no matter what those words are. This creates at least one more vector for coin loss: accidentally entering your password instead of the 12 word seed phrase. Even though the password is not in the mnemonic word list, this will not get checked because of the use of range(len(wlist)/3).