BinaryInfo should not be a functor

[edsko]

At the moment we have

data BinaryInfo blob = BinaryInfo
  { binaryBlob   :: !blob
  , headerOffset :: !Word16
    -- ^ The offset within the 'binaryBlob' at which the header starts.
  , headerSize   :: !Word16
    -- ^ How many bytes the header is long. Extracting the 'headerSize' bytes
    -- from 'binaryBlob' starting from 'headerOffset' should yield the header.

    -- In the future, i.e. Shelley, we might want to extend this to include a
    -- field to tell where the transaction body ends and where the transaction
    -- witnesses begin so we can only extract the transaction body.
  }

but that feels wrong: there is no way that a Functor instance could preserve any kind of internal invariants of this type, and even if this isn't true in general, I suspect even for specific use cases of fmap those invariants are very unlikely to be preserved.