Clarify enr content value format

[mattsse]

This states:

https://github.com/sigp/enr/blob/83cd98990e7f191b310addb202869994a8f59c96/src/builder.rs#L18-L20

This will add any encodable value encoded as rlp to the map:

https://github.com/sigp/enr/blob/83cd98990e7f191b310addb202869994a8f59c96/src/builder.rs#L45-L56

on build this check ensures that all values are rlp(bytes):

https://github.com/sigp/enr/blob/83cd98990e7f191b310addb202869994a8f59c96/src/builder.rs#L159-L162

on reth, this test currently fails because the value is rlp(forkId)

https://github.com/paradigmxyz/reth/blob/bc2634981e2fdca2d968b1dc63d3ca9a716c93e4/crates/net/discv4/src/proto.rs#L660-L664

If all values should be rlp(bytes) where bytes itself can be rlp(forkid), should add_value then insert self.add_value_rlp(key, alloy_rlp::encode(&out)) ?

https://github.com/sigp/enr/blob/83cd98990e7f191b310addb202869994a8f59c96/src/builder.rs#L45-L50

[AgeManning]

To be generic across all types we store the underlying data as bytes and convert to specific types later one.

The general spec indicates an RLP list of k,v: https://github.com/ethereum/devp2p/blob/master/enr.md#rlp-encoding

In principle, it should work for any object that is RLP encodable. In your test could you not just put the fork_id into add_value() and it will be encoded as every other type that is rlp encodeable and stored in the ENR.

Your suggestion would be to double rlp-encode objects right? Currently we are just rlp-encoding the types and storing the bytes of the rlp-encoded object, which seems sensible to me.

Unless the .build() fails on normal types, it should be the case that all rlp-bytes can be decoded as bytes right?