Use borsh instead of SCALE within tendermint-machine, tributary-sdk

Not only does this follow our general practice, the latest SCALE has a
possibly-lossy truncation in its current implementation for `enum`s I'd like to
avoid without simply silencing.

coordinator/tributary-sdk/src/tendermint/tx.rs

@@ -1,6 +1,6 @@
 use std::io;
 
-use scale::{Encode, Decode, IoReader};
+use borsh::BorshDeserialize;
 
 use blake2::{Digest, Blake2s256};
 
@@ -27,14 +27,14 @@ pub enum TendermintTx {
 
 impl ReadWrite for TendermintTx {
   fn read<R: io::Read>(reader: &mut R) -> io::Result<Self> {
-    Evidence::decode(&mut IoReader(reader))
+    Evidence::deserialize_reader(reader)
       .map(TendermintTx::SlashEvidence)
       .map_err(|_| io::Error::new(io::ErrorKind::InvalidData, "invalid evidence format"))
   }
 
   fn write<W: io::Write>(&self, writer: &mut W) -> io::Result<()> {
     match self {
-      TendermintTx::SlashEvidence(ev) => writer.write_all(&ev.encode()),
+      TendermintTx::SlashEvidence(ev) => writer.write_all(&borsh::to_vec(&ev).unwrap()),
     }
   }
 }