Convert Jint.Benchmark to use BenchmarkDotNet

[lahma]

While discussing a performance problem when using RavenDB I also checked some Jint code. There might be possibilities for performance optimizations but there should be solid numbers if changes are made. So I converted the current benchmark project to use BenchmarkDotNet.

Now we can get better numbers and memory allocations from running the Jurassic vs Jint performance test in GitHub pasteable format by running dotnet run -c Release in benchmark project directory:

BenchmarkDotNet=v0.10.11, OS=Windows 10 Redstone 3 [1709, Fall Creators Update] (10.0.16299.98)
Processor=Intel Core i7-6820HQ CPU 2.70GHz (Skylake), ProcessorCount=8
Frequency=2648437 Hz, Resolution=377.5812 ns, Timer=TSC
.NET Core SDK=2.1.2
  [Host]     : .NET Core 2.0.3 (Framework 4.6.25815.02), 64bit RyuJIT
  DefaultJob : .NET Core 2.0.3 (Framework 4.6.25815.02), 64bit RyuJIT

Method	Iterations	ReuseEngine	Mean	Error	StdDev	Gen 0	Gen 1	Allocated
Jint	10	False	1,766.2 us	8.986 us	7.504 us	453.1250	5.8594	1858.52 KB
Jurassic	10	False	51,225.1 us	647.867 us	574.317 us	1062.5000	375.0000	6057 KB
Jint	10	True	960.7 us	14.484 us	13.548 us	125.9766	-	519.06 KB
Jurassic	10	True	49,007.5 us	1,217.638 us	2,132.596 us	62.5000	-	450.25 KB
Jint	20	False	3,647.4 us	43.487 us	38.550 us	906.2500	11.7188	3717.03 KB
Jurassic	20	False	103,830.2 us	1,997.076 us	2,927.289 us	2187.5000	750.0000	12113.99 KB
Jint	20	True	1,920.6 us	37.992 us	58.017 us	251.9531	-	1038.13 KB
Jurassic	20	True	109,151.1 us	2,157.805 us	5,043.800 us	187.5000	62.5000	900.53 KB

[sebastienros]

Next step is to have benchmarks that spread different parts of the codebase. Like the parser, Arrays, ...or even Sunspider scripts.

[lahma]

I could gladly try to create at least benchmark case for running script against large nested object structure, it showed some places to improve where caching was not present.