Consume Roslyn with ref fields support

[AaronRobinsonMSFT]

Moves to a Roslyn version that support ref field support and the scoped keyword.

Removes ByReference<T> from all runtimes.

Contributes to #63768

[dotnet-issue-labeler]

Note regarding the new-api-needs-documentation label:

This serves as a reminder for when your PR is modifying a ref *.cs file and adding/modifying public APIs, to please make sure the API implementation in the src *.cs file is documented with triple slash comments, so the PR reviewers can sign off that change.

[dotnet-issue-labeler]

I couldn't figure out the best area label to add to this PR. If you have write-permissions please help me learn by adding exactly one area label.

[ghost]

Tagging subscribers to this area: @dotnet/area-meta
See info in area-owners.md if you want to be subscribed.

Issue Details

---

Moves to a Roslyn version that support ref field support and the scoped keyword.

Removes ByReference<T> from all runtimes.

Author:	AaronRobinsonMSFT
Assignees:	AaronRobinsonMSFT
Labels:	area-Meta, CoreClr, Mono
Milestone:	-

[AaronRobinsonMSFT]

@AaronRobinsonMSFT I'm not seeing any interpreter issues on the tests. Is there anything else that needs fixing ?

@BrzVlad I think we've made it past most of the legs that were failing so I perhaps this is solved. I applied your suggestion and verified it locally fixed the issue I was hitting above - Thank you.

@BrzVlad or @lambdageek Can I get one of you to sign-off on the mono changes?

[EgorBo]

public Span<byte> Test(byte[] array) => array.AsSpan(0, 10);

codegen before this PR:

; Method Program:ArrayAsSpanStartLength(System.Byte[]):System.Span`1[Byte]:this
G_M63029_IG01:      
       4883EC28             sub      rsp, 40

G_M63029_IG02:            
       4D85C0               test     r8, r8
       741D                 je       SHORT G_M63029_IG04
       418378080A           cmp      dword ptr [r8+8], 10
       7216                 jb       SHORT G_M63029_IG04
       4983C010             add      r8, 16
       4C8902               mov      bword ptr [rdx], r8
       C742080A000000       mov      dword ptr [rdx+8], 10
       488BC2               mov      rax, rdx

G_M63029_IG03:         
       4883C428             add      rsp, 40
       C3                   ret      

G_M63029_IG04:          
       FF15247D2700         call     [System.ThrowHelper:ThrowArgumentOutOfRangeException()]
       CC                   int3     
; Total bytes of code: 45

Codegen after this PR:

; Method Program:ArrayAsSpanStartLength(System.Byte[]):System.Span`1[Byte]:this
G_M63029_IG01:  
       57                   push     rdi
       56                   push     rsi
       53                   push     rbx
       4883EC30             sub      rsp, 48
       C5F877               vzeroupper 
       488BDA               mov      rbx, rdx

G_M63029_IG02:             
       C5F857C0             vxorps   xmm0, xmm0
       C5FA7F442420         vmovdqu  xmmword ptr [rsp+20H], xmm0
       4D85C0               test     r8, r8
       7432                 je       SHORT G_M63029_IG04
       418378080A           cmp      dword ptr [r8+8], 10
       722B                 jb       SHORT G_M63029_IG04
       4983C010             add      r8, 16
       4C89442420           mov      bword ptr [rsp+20H], r8
       C74424280A000000     mov      dword ptr [rsp+28H], 10
       488BFB               mov      rdi, rbx
       488D742420           lea      rsi, bword ptr [rsp+20H]
       E8DF738B5F           call     CORINFO_HELP_ASSIGN_BYREF
       48A5                 movsq    
       488BC3               mov      rax, rbx

G_M63029_IG03:         
       4883C430             add      rsp, 48
       5B                   pop      rbx
       5E                   pop      rsi
       5F                   pop      rdi
       C3                   ret      

G_M63029_IG04:
       FF158CDB0B00         call     [System.ThrowHelper:ThrowArgumentOutOfRangeException()]
       CC                   int3     
; Total bytes of code: 85

perf-triage team (@DrewScoggins, @tannergooding, @kunalspathak) noticed that this PR caused a lot of perf regressions

[tannergooding]

Notably SuperPMI says "zero diffs" which is interesting. Possibly two changes not playing well together or maybe something SuperPMI doesn't catch?

[kunalspathak]

or maybe something SuperPMI doesn't catch?

@BruceForstall

[jakobbotsch]

Notably SuperPMI says "zero diffs" which is interesting. Possibly two changes not playing well together or maybe something SuperPMI doesn't catch?

Since all the significant changes here are on the VM side that is to be expected. SPMI will only reflect JIT changes.

[DrewScoggins]

We are also seeing over 10% regressions in plaintext TE