Some panics to be handled on adversarial input

[ebuchman]

No need to merge this. Probably we should start an Adversary folder in the Tests directory and begin to standardize this sort of thing. But here I'm pointing out that

• NewValueFromBytes panics on some forms of bad rlp
• ValidateTransaction panics on a bad signature (this is fixed here, by making the sig check explicit: fix panic on bad tx sender #195)

These are important because they are called more or less directly after receiving raw bytes from peers.

[fjl]

Uses of ethutil.Decode (and hopefully also ethutil.Value) will be replaced by package rlp. I'm working on it right now.

We are also working on a new peer to peer networking layer in package p2p (see #155). It will be a lot less racy.

[ebuchman]

Ok awesome. Am looking forward to it. Just trying to avoid peers being able to crash clients by sending bad data.

As for the txpool, it's simply missing a sig check (rather than being rlp related). I will add the test to my other PR fixing it (#195), I simply wanted to avoid duplicating the fakeEth which is also present in #193, but now I think maybe it will be necessary to duplicate, as they serve different needs.

[obscuren]

Please re-open if you want to work in the new rlp decoder/encoder