Environment.ProcessorCount does not take into account cgroup/Docker limits

[tmds]

This leads to managed code over-dimensioning when running in Docker.

coreclr takes into account docker cpu limits: https://github.com/dotnet/coreclr/blob/dbfd10c94055fb5af823b4a7e30774641afcb3b2/src/classlibnative/bcltype/system.cpp#L349-L350

[terrajobst]

@stephentoub @weshaggard

[tmds]

@stephentoub @janvorli I'd like to do a PR for this, but I'm not sure how I should call the PAL_GetCpuLimit of coreclr in corefx?

[jkotas]

I think we should switch Environment.ProcessorCount to delegate to the runtime:

• Add ProcessorCount to https://github.com/dotnet/coreclr/blob/master/src/mscorlib/src/Internal/Runtime/Augments/EnvironmentAugments.cs and https://github.com/dotnet/corert/blob/master/src/System.Private.CoreLib/src/Internal/Runtime/Augments/EnvironmentAugments.Unix.cs / https://github.com/dotnet/corert/blob/master/src/System.Private.CoreLib/src/Internal/Runtime/Augments/EnvironmentAugments.Windows.cs.
• Change the CoreFX implementation to forward ProcessorCount to the runtime.

[tmds]

Thanks @jkotas I will take a shot at implementing that.