[JIT] ARM64 - Fixed regressions for GT_NEG containment

[TIHan]

Resolves #85225

[ghost]

Tagging subscribers to this area: @JulieLeeMSFT, @jakobbotsch
See info in area-owners.md if you want to be subscribed.

Issue Details

---

Resolves #85225

Author:	TIHan
Assignees:	TIHan
Labels:	area-CodeGen-coreclr
Milestone:	-

[TIHan]

@dotnet/jit-contrib @jakobbotsch this is ready

[jakobbotsch]

/azp run Fuzzlyn

[azure-pipelines]

Azure Pipelines successfully started running 1 pipeline(s).

[jakobbotsch]

Looks like there's still some examples, e.g.:

// Generated by Fuzzlyn v1.5 on 2023-04-24 10:09:34
// Run on Arm64 Linux
// Seed: 15668805006347923926
// Reduced from 103.1 KiB to 0.3 KiB in 00:01:18
// Debug: Outputs False
// Release: Outputs True
public class Program
{
    public static long s_7;
    public static int[] s_11 = new int[]{0};
    public static void Main()
    {
        s_11[0] = -2147483648;
        var vr9 = (int)s_7 < (-s_11[0]);
        System.Console.WriteLine(vr9);
    }
}

[TIHan]

Updated again, this should fix it for good now. I was containing NEG for all comparison ops, but it's only valid for EQ and NE.

[jakobbotsch]

/azp run Fuzzlyn

[azure-pipelines]

Azure Pipelines successfully started running 1 pipeline(s).

[TIHan]

/azp run Fuzzlyn

[azure-pipelines]

Azure Pipelines successfully started running 1 pipeline(s).

[TIHan]

CI failure is unrelated. Last Fuzzlyn runs succeeded, though it did catch an unrelated case in OSX which I reported here.