zero-pad signature bytes of messages, return r & s as integers

[carver]

What was wrong?

#466 signature bytes were not 0-padded on message signatures

How was it fixed?

• 0-pad them r and s in signature bytes
• return r and s as integers
• r and s in transactions remain not 0-padded

Cute Animal Picture

[pipermerriam]

I'm thinking that a hypothesis test for this might be a good thing since we have some reasonably strong properties that we can assert about signatures and signed transactions and that type of test likely would have caught this bug.

• A signed transaction's signature recovers to the expected sender address
• The raw transaction can be rlp decoded into a transaction object
• An rlp decoded transaction object's v,r,s match the v,r,s from the signature.
• The signature bytes are 65 bytes in length

Those are the ones that come to mind.

[pipermerriam]

r_bytes, s_bytes = to_bytes32(r), to_bytes32(s)
# or
r_bytes = to_bytes32(r)
s_bytes = to_bytes32(s)

Minor readability improvements?