[C#] Make DirectBuffer.CheckLimit inlinable

[szehetner]

DirectBuffer.CheckLimit(int limit) is called in a lot of places during encoding and decoding and would profit from being inlined by the JIT Compiler. Currently it can't because of the throw statements and it's size. This change extracts the cold path where the check fails into a separate method, making CheckLimit itself eligible for inlining and resulting in a measureable performance improvement.

Running the existing benchmark project:
Before:

##teamcity[buildStatisticValue key='AverageEncodeLatencyNanos' value='40,0']
##teamcity[buildStatisticValue key='AverageDecodeLatencyNanos' value='33,8']

After:

##teamcity[buildStatisticValue key='AverageEncodeLatencyNanos' value='33,6']
##teamcity[buildStatisticValue key='AverageDecodeLatencyNanos' value='28,4']

I also tried a BenchmarkDotNet benchmark based on the car example (https://gist.github.com/szehetner/5ed4f7acceedfb8bce092b10f829ba92)

BenchmarkDotNet=v0.12.1, OS=Windows 10.0.18363.959 (1909/November2018Update/19H2)
Intel Core i9-10885H CPU 2.40GHz, 1 CPU, 16 logical and 8 physical cores
.NET Core SDK=5.0.200
  [Host]     : .NET Core 5.0.3 (CoreCLR 5.0.321.7212, CoreFX 5.0.321.7212), X64 RyuJIT
  DefaultJob : .NET Core 5.0.3 (CoreCLR 5.0.321.7212, CoreFX 5.0.321.7212), X64 RyuJIT

before:

Method	Mean	Error	StdDev	Gen 0	Gen 1	Gen 2	Allocated
Encode	189.5 ns	3.77 ns	3.70 ns	-	-	-	-
Decode	167.9 ns	1.67 ns	1.39 ns	-	-	-	-

after:

Method	Mean	Error	StdDev	Gen 0	Gen 1	Gen 2	Allocated
Encode	158.9 ns	1.93 ns	1.80 ns	-	-	-	-
Decode	129.5 ns	0.57 ns	0.48 ns	-	-	-	-

Btw, I would suggest to convert the existing benchmark project to BenchmarkDotNet, as this is the de-facto standard for .NET benchmarks. I could create another PR for this if desired.

[billsegall]

Looks good. That's a substantial improvement.

A separate PR for BenchmarkDotNet sounds great.

Thanks!