Added a TypeRegistrar to CommandAppTester

[nils-a]

exposed as a public property.
For convenience, and to keep the similarity to the real CommandApp
it is also available in the ctor of CommandAppTester.

This is in regard to #461

I'm not quite fond of the used-in-this-one-place implementation of FakeRegistrar, but I see no real good alternative except exposing the DefaultRegistrar to be used in tests or creating a second (fully implemented) ITypeRegistrar for tests.

[CLAassistant]

All committers have signed the CLA.

[patriksvensson]

@nils-a I just pushed a fix for the analyzer where we moved the test project to the root. Sorry, didn't think about your PR when I did that so you will have to rebase your branch and fix the merge conflict 😢

I think the fake registrar is OK but perhaps we could re-use the FakeTypeRegistrar for this purpose?

[nils-a]

@patriksvensson

so you will have to rebase your branch and fix the merge conflict

Not a problem. 👍

perhaps we could re-use the FakeTypeRegistrar for this purpose?

I've thought about that. (In fact I implemented it, then reverted..) The problem I see with that approach is that FakeTypeRegistrar currently returns null on Build(). Implementing an ITypeResolver would be a breaking change in Spectre.Console.Testing.

But... Thinking about it now... I could add the ITypeResolver as optional feature. Have the result of the Build method injectable per Func.

[patriksvensson]

LGTM!