Optimise CBOR-in-CBOR (de)serialisation

The BlockFetch and ChainSync codecs both use CBOR-in-CBOR.
Encoding and decoding could be more efficient if we had `encodeEmbeddedCBOR` ([1]) ([2]) and `decodeEmbeddedCBOR` ([3]) ([4]).

Additionally, we could benefit from having a variant of `encodeBytes` that works for lazy bytestrings (e.g., `encodeBytesChunked`). In case the lazy bytestring contains multiple chunks, it first needs to be copied into one contiguous strict bytestring before it can be passed to `encodeBytes`. Avoiding that needless copy would be nice ([5]).

[1]: https://github.com/input-output-hk/ouroboros-network/blob/971b0b58d7833ac358aa8285ae6c58a00c3f22b9/ouroboros-network/src/Ouroboros/Network/Protocol/BlockFetch/Codec.hs#L94
[2]: https://github.com/input-output-hk/ouroboros-network/blob/971b0b58d7833ac358aa8285ae6c58a00c3f22b9/ouroboros-network/src/Ouroboros/Network/Protocol/ChainSync/Codec.hs#L135
[3]: https://github.com/input-output-hk/ouroboros-network/blob/971b0b58d7833ac358aa8285ae6c58a00c3f22b9/ouroboros-network/src/Ouroboros/Network/Protocol/BlockFetch/Codec.hs#L101
[4]: https://github.com/input-output-hk/ouroboros-network/blob/971b0b58d7833ac358aa8285ae6c58a00c3f22b9/ouroboros-network/src/Ouroboros/Network/Protocol/ChainSync/Codec.hs#L142
[5]: https://github.com/input-output-hk/ouroboros-network/pull/1271/files#r353642273