Investigate improving DI container runtime codegen

[davidfowl]

We want to see if we can improve the performance of our DI container by looking at a few different things:

• Reducing locks when resolving singletons/scoped services
• Improving accessing fields of the closure passed into the compiled lambda expression. Today each access is an array access. It can be a field access (see https://github.com/dadhi/FastExpressionCompiler)

We also want to look at adding the ability to pre-compile the DI container to avoid any runtime cost at all. This is obviously tricky for things like open generics but we can fallback to a runtime implementation where possible.

/cc @pakrym

[dadhi]

@davidfowl

We also want to look at adding the ability to pre-compile the DI container to avoid any runtime cost at all.

I had a probably related question.

[dzmitry-lahoda]

@davidfowl, I also want to see improvements in areas meantioned. But do you have any data that these improvements are important? What area is more important than other?

[aspnet-hello]

This issue was moved to dotnet/aspnetcore#2338