Initial benchmarks of wasm contracts compared to sdk modules

[ethanfrey]

Starts on #635

Get some initial data points

[codecov]

Codecov Report

Merging #638 (13aebec) into master (dcb0425) will increase coverage by 0.03%.
The diff coverage is n/a.

@@            Coverage Diff             @@
##           master     #638      +/-   ##
==========================================
+ Coverage   60.21%   60.24%   +0.03%     
==========================================
  Files          48       48              
  Lines        5313     5313              
==========================================
+ Hits         3199     3201       +2     
+ Misses       1889     1888       -1     
+ Partials      225      224       -1     

Impacted Files	Coverage Δ
app/test_helpers.go	0.86% <ø> (ø)
x/wasm/keeper/keeper.go	87.30% <0.00%> (+0.33%)	⬆️

[ethanfrey]

This used memdb and 20tx/block and some unrealistic settings used in the cosmos-sdk benchmarks.

See initial results

goos: linux
goarch: amd64
pkg: github.com/CosmWasm/wasmd/benchmarks
cpu: Intel(R) Xeon(R) Platinum 8124M CPU @ 3.00GHz
BenchmarkNCw20SendTxPerBlock-36     	     832	   1549412 ns/op
BenchmarkNBankSendTxsPerBlock-36    	     943	   1291934 ns/op
PASS
ok  	github.com/CosmWasm/wasmd/benchmarks	3.624s

Cw20Send is about 20% slower than Bank Send (including signature verification, ante handlers, etc)