Sve2_ConvertToSingleOddRoundToOdd_float_double test failure

[jakobbotsch]

Example pipeline run: https://dev.azure.com/dnceng-public/public/_build?definitionId=111&_a=summary
Example console log: https://helixr18s23ayyeko0k025g8.blob.core.windows.net/dotnet-runtime-refs-heads-main-6ad8f9e9eaf347ec82/HardwareIntrinsics_Arm_ro/1/console.7125161f.log

18:48:42.630 Running test: _Sve2_ro::JIT.HardwareIntrinsics.Arm._Sve2.Program.Sve2_ConvertToSingleOddRoundToOdd_float_double()
Beginning scenario: RunBasicScenario_UnsafeRead
Beginning scenario: RunBasicScenario_Load
Beginning scenario: RunReflectionScenario_UnsafeRead
Beginning scenario: RunLclVarScenario_UnsafeRead
Beginning scenario: RunSameLclVarScenario_UnsafeRead
Beginning scenario: RunClassFldScenario
Beginning scenario: RunSameClassFldScenario
Beginning scenario: RunStructLclFldScenario
Beginning scenario: RunStructFldScenario
Beginning scenario: ConditionalSelect_Op1_mask - operation in TrueValue
Beginning scenario: ConditionalSelect_Op1_zero - operation in TrueValue
Beginning scenario: ConditionalSelect_Op1_all - operation in TrueValue
Beginning scenario: ConditionalSelect_Op1_mask - operation in FalseValue
Beginning scenario: ConditionalSelect_Op1_zero - operation in FalseValue
Beginning scenario: ConditionalSelect_Op1_all - operation in FalseValue
Beginning scenario: ConditionalSelect_Op2_mask - operation in TrueValue
Beginning scenario: ConditionalSelect_Op2_zero - operation in TrueValue
Sve2.ConvertToSingleOddRoundToOdd<Single>(Vector<Single>, Vector<Double>): ConditionalSelectScenario_TrueValue failed:
    mask: (0, 0, 0, 0)
    left: (0.8978391, 0.6871079, 0.08343614, 0.1541698)
   right: (0.7538663866339904, 0.5267524715264288)
 falseOp: (3.133143E+23, 1.8134665, NaN, 1.756688)
  result: (3.133143E+23, 1.8134665, NaN, 1.756688)

Introduced by #118957? cc @SwapnilGaikwad

[dotnet-policy-service]

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

[SwapnilGaikwad]

Thanks, @jakobbotsch. It's likely due to incorrect NaN comparisons and mentioned #121659. The fix (#121706) is ready for review. If this issue persists after the fix, we can revisit.