Library is not flexible enough to handle Parity's custom compression

[lithp]

Parity compresses and decompresses the rlp objects which it writes into the database. It has a hand-selected list of substitutions which it applies to strings (not lists). Attempting to rlp.decode one of the compressed objects results in rlp.exceptions.DecodingErrors. This is a good thing, because the compressed objects aren't valid rlp!

However, because of the way in which pyrlp is written it is very difficult to use pyrlp for reading these objects. consume_payment tries to recurse into rlp objects and doesn't provide any extension points for swapping out \x81\x02 (an example of an invalid sequence which appears in the compressed object) with the decompressed value. consume_length_prefix isn't usable either because it throws an exception on these objects, in order to read these objects I needed to implement my own.

I think this could be fixed by moving some exceptions around and breaking up a few functions such that users could consume just the logic they needed.