Moved CodeGen/EmittedIL/LiteralValue over to NUnit

[sergey-tihon]

part of #7075

[sergey-tihon]

@TIHan I have to add module line and rename arg to _ to make this test works.
It is OK that CompilerAssert.CompileLibraryAndVerifyIL require source code to compile without warnings?

[sergey-tihon]

In general I do not like that we remove comments from emitted IL code

fsharp/tests/fsharp/Compiler/ILChecker.fs

Lines 46 to 54 in 8209f2f

	let textNoComments =
	System.Text.RegularExpressions.Regex.Replace(text,
	blockComments + "|" + lineComments + "|" + strings + "|" + verbatimStrings,
	(fun me ->
	if (me.Value.StartsWith("/*") || me.Value.StartsWith("//")) then
	if me.Value.StartsWith("//") then Environment.NewLine else String.Empty
	else
	me.Value), System.Text.RegularExpressions.RegexOptions.Singleline)
	|> filterSpecialComment

and do not do it for expectedIL samples.

In this test I have to remove all comments manually.
Can we apply the same transformation to emitted and expected IL? What do think @TIHan ?

[cartermp]

This is necessary since some tools (ILDasm?) add comments such that you cannot verify IL emitted by the compiler. This was necessary for string concat optimization work I believe.

[sergey-tihon]

@cartermp I got it, but I think that would be nice to allow copy generated IL from ILDasm and use it in test directly (without manual comments removal).

Then test framework (inside verifier.VerifyIL):

1. compile code sample and generate IL using ILDASM
2. remove comments from IL produced in step 1 (like it does now)
3. remove comments from expected IL specified in using test, the same transformation that we do in step 2 (I propose)
4. compare two set of IL instructions from 2 & 3

benefits: less manual work to cleanup expected IL (ability to use snippet from ILDasm).

[sergey-tihon]

This PR should be ready for review (if tests pass).
My questions/concerns:

1. I have to update source code to compile without warnings to run test (I think that IL tests should not require this)
2. I think that VerifyIL should remove comments from emitted and expected IL (apply same transformation for both IL samples)
3. Would be nice to make the error message more readable. Maybe it is personal, but I think if it test fail then It is hard to understand where the difference is. Ok, in the console with reasonable expectedIL message is readable
4. Please review expected IL, I've used only part of original file and slightly modified it. How much IL do you want to keep in expected IL? My understanding that .field public static literal int32 x = int32(0x00000007) would be enough for this case.

[sergey-tihon]

Because of these errors on corecrl I decided to minimize expected IL

==
Name: .class public abstract auto ansi sealed LiteralValue

Expected:	 extends [mscorlib]System.Object
Actual:		 extends [System.Runtime]System.Object
==

==
Name: .class public abstract auto ansi sealed LiteralValue

Expected:	 .custom instance void [mscorlib]System.Diagnostics.DebuggerBrowsableAttribute::.ctor(valuetype [mscorlib]System.Diagnostics.DebuggerBrowsableState) = ( 01 00 00 00 00 00 00 00 )
Actual:		 .custom instance void [System.Diagnostics.Debug]System.Diagnostics.DebuggerBrowsableAttribute::.ctor(valuetype [System.Diagnostics.Debug]System.Diagnostics.DebuggerBrowsableState) = ( 01 00 00 00 00 00 00 00 )
==