codec: new TxHandler byte decoder

[algorandskiy]

Summary

Use msgp marshaller instead of reflect in TxHandler

BenchmarkTxHandlerDecoder-8       	      68	  17116004 ns/op	 3335105 B/op	   51105 allocs/op
BenchmarkTxHandlerDecoderMsgp-8   	     122	  10213123 ns/op	 1189860 B/op	     614 allocs/op

BenchmarkTxHandlerProcessDecoded-8   	   62826	     19184 ns/op	    8956 B/op	      57 allocs/op
vs old
BenchmarkTxHandlerProcessDecoded-8   	   47414	     21560 ns/op	    8957 B/op	      57 allocs/op

Test Plan

TODO: write tests

[codecov]

Codecov Report

Merging #4266 (1a12135) into master (bf7aaec) will increase coverage by 0.03%.
The diff coverage is 42.30%.

@@            Coverage Diff             @@
##           master    #4266      +/-   ##
==========================================
+ Coverage   55.18%   55.22%   +0.03%     
==========================================
  Files         398      398              
  Lines       50137    50148      +11     
==========================================
+ Hits        27669    27692      +23     
+ Misses      20156    20141      -15     
- Partials     2312     2315       +3     

Impacted Files	Coverage Δ
cmd/goal/clerk.go	8.75% <0.00%> (ø)
cmd/goal/multisig.go	10.00% <0.00%> (ø)
protocol/codec_tester.go	10.61% <0.00%> (ø)
data/txHandler.go	11.11% <20.00%> (+11.11%)	⬆️
protocol/codec.go	64.21% <80.00%> (+1.85%)	⬆️
cmd/tealdbg/local.go	73.95% <100.00%> (ø)
ledger/roundlru.go	90.56% <0.00%> (-5.67%)	⬇️
ledger/blockqueue.go	84.48% <0.00%> (-1.15%)	⬇️
data/transactions/verify/txn.go	43.85% <0.00%> (-0.88%)	⬇️
ledger/acctupdates.go	69.29% <0.00%> (-0.60%)	⬇️
... and 8 more

📣 We’re building smart automated test selection to slash your CI/CD build times. Learn more

[jannotti]

It seems like this allows for every single call to protocol.NewDecoderBytes([]byte) to be replaced by your new NewMsgpDecoderBytes([]byte) unless the returned decoder is being used in a weird way. Seems like all we do with them is call Decode. And if we know we are decoding msgpack data (implied by the fact we set the codec to msgp in NewDecoderBytes), there's no point in using the Decoder type and going through reflection in Decoder.Decode() just let the Decode() on MsgpDecoderBytes get the job done statically.

[algorandskiy]

Thank you for the feedback, I'll add a protocol/codec random data/length unit test and merge.

[algorandskiy]

Rebased/resolved a merge conflict