specialized checksumPrepend and checksumAppend

[nikiforo]

The current implementation of checksummed implies that the developer knows framing. Though, sometimes it's easier to define a codec by providing the target Codec and the parameters for the checksum, e.g. size and an algorithm. For instance, look at roundtrip undefined size, I'm not sure this logic can be easily expressed using checksummed to append the CRC.

From my POV, the construction of ChecksumPrepend in terms of checksummed is harder than it should be. IIRC, I tried to express it with Codec.exmap(_.swap)(_.swap) but it was too difficult to understand when I looked at it later.

Moreover, framing in checksummed controls not only the place of CRC value and the region of bits to compute the CRC but also what is being passed to the target. That not only defines how the CRC is computed, but also what is fed to the target, which seems to be wrong, at least by violating the SRP. Also, framing doesn't allow a user to omit bit regions for CRC computation. I encountered it when I implemented a codec with the structure | THE_CRC | REGION NOT UNDER CRC | REGION UNDER THE_CRC |