Decode msgpack and json consistently

[eddyashton]

Our partially recursive failed to decode strings in lists and key-values, causing inconsistent behaviour depending on whether the format chosen was "json" or "msgpack".

>>> o
{'abc': 'def', 'ghi': {'jkl': ['a', 'b', 'c'], 'mno': {}}}
>>> json.loads(json.dumps(o))
{'abc': 'def', 'ghi': {'jkl': ['a', 'b', 'c'], 'mno': {}}}
>>> msgpack.unpackb(msgpack.packb(o))
{b'abc': b'def', b'ghi': {b'jkl': [b'a', b'b', b'c'], b'mno': {}}}

Ideally they would be identical, and they are if we call msgpack.unpackb(..., raw=False).

As an associated change, the raw quote is passed as a byte-vector rather than a string. This does mean it appears as a long int list when looking at the raw JSON/Python, so its a prime candidate to be hex-encoded (#278).

[cimetrics]

[codecov-io]

Codecov Report

Merging #566 into master will increase coverage by 0.22%.
The diff coverage is 0%.

@@            Coverage Diff             @@
##           master     #566      +/-   ##
==========================================
+ Coverage   78.42%   78.64%   +0.22%     
==========================================
  Files         140      140              
  Lines       10508    10508              
==========================================
+ Hits         8240     8263      +23     
+ Misses       2268     2245      -23

Flag	Coverage Δ
#e2e_BFT	50.28% <0%> (-0.45%)	⬇️
#e2e_CFT	73.43% <0%> (+0.11%)	⬆️
#unit_BFT	64.39% <ø> (-0.83%)	⬇️
#unit_CFT	72.74% <ø> (ø)	⬆️

Impacted Files	Coverage Δ
src/node/rpc/nodecalltypes.h	100% <ø> (ø)	⬆️
src/node/nodestate.h	83.48% <0%> (ø)	⬆️
src/tls/intel_drng.h	72% <0%> (+0.8%)	⬆️
src/host/tcp.h	73.93% <0%> (+0.85%)	⬆️
src/consensus/raft/raft.h	83.86% <0%> (+1.27%)	⬆️
src/tls/entropy.h	94.12% <0%> (+70.59%)	⬆️

[cimetrics]