Replace bincode with borsh

[kayabaNerve]

Replaces a non-canonical encoding with a canonical encoding which additionally may be faster slightly faster, yet at least is effectively the same.

Also fixes multiple issues such as:

• Usage of bincode (a non-canonical format) in transcripts
• create_db! not binding to type in set, correction of which raised further errors

And causes improvements such as:

• No-longer JSON encoding processor messages in a nested byte buffer. A further PR should move message-queue off HTTP JSON-RPC to raw sockets entirely.
• Not needing a Vec for some processor messages, as we can now use [u8; 64]
• 10-20% reduced disk usage, per generic benchmark results present in https://github.com/djkoloski/rust_serialization_benchmark.

While this was initially meant to be an experiment like the QUIC branch, the observed issues and included fixes necessitate merging of this branch or porting it to a new PR and merging that.