Audit error types propagated from p2p client to checkpoint sync

[CHr15F0x]

Any errors coming from p2p::peer_agnostic::Client::*_stream should be irrecoverable errors.

For example:

• class_definition_stream returns a fatal error (wrapped in anyhow::Error) only when the DB fails, because it ingests a stream of expected class counts per block from the DB.

[CHr15F0x]

More concrete steps:

• find recoverable errors propagated as anyhow::Error
• introduce specific error types for those :arrow_up:
• rename the Other variant to Fatal
• remove one of: (SyncError, SyncError2) OR at least make sure that shared_impls (like in class_definitions.rs) do not return anyhow::Error but MySpecificErrorNewtype(anyhow::Error|innerType) which then forces map_err to correct SyncError|SyncError2 variant and cannot be by mistake implicitly converted to the ::Fatal() variant of SyncError|SyncError2
• add meaningful data to the error variants (ie. which block, which class, which txn, specific error details, etc.)