Do not change abstract class ctors to protected

[latkin]

Fix for #420

[latkin]

Ah crud, didn't think to run core tests. Will update that...

[dsyme]

Hmmm... @latkin - I wonder why we emitted "protected" on constructors for abstract classes though - there must have been some reason why we do that. Does C# really never do it?

thanks
don

[latkin]

I chatted w/ @VSadov yesterday and he says C# never alters the accessibility in this case. Quick peek at compiled assembly confirms.

I don't know if F# compiler makes some assumptions elsewhere that rely on conversion to protected. Compiler tests ran ok with the change.

    public abstract class A
    {
        internal A(string x) { }
        public A(int x) { }
        private A(float x) { }
        protected A(double x) { }
        internal protected A(byte x) { }
    }

.class public auto ansi abstract beforefieldinit CSConsoleApp1.A
    extends [mscorlib]System.Object
{
    // Methods
    .method assembly hidebysig specialname rtspecialname 
        instance void .ctor (
            string x
        ) cil managed 
    { ... }

    .method public hidebysig specialname rtspecialname 
        instance void .ctor (
            int32 x
        ) cil managed 
    { ... }

    .method private hidebysig specialname rtspecialname 
        instance void .ctor (
            float32 x
        ) cil managed 
    { ... }

    .method family hidebysig specialname rtspecialname 
        instance void .ctor (
            float64 x
        ) cil managed 
    { ... }

    .method famorassem hidebysig specialname rtspecialname 
        instance void .ctor (
            uint8 x
        ) cil managed 
    { ... }

} // end of class CSConsoleApp1.A

[latkin]

Updated core surface area tests, things are green now

[dsyme]

LGTM 👍