Add unsigned variants for BitConverter float bit APIs

[MichalPetryka]

Adds DoubleToUInt64Bits, UInt64BitsToDouble, SingleToUInt32Bits,
UInt32BitsToSingle, HalfToUInt16Bits and UInt16BitsToHalf.

Implementations were based on existing signed integer variants.

Convert usages of existing APIs to unsigned variants when appropriate.

Fix #36469

The changes build on my machine, but I was unable to run the unit tests, running all of them made the test process hang for 4 hours, so I'm opening this to see what happens on the CI.

[ghost]

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.

[ghost]

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

Issue Details

---

Adds DoubleToUInt64Bits, UInt64BitsToDouble, SingleToUInt32Bits,
UInt32BitsToSingle, HalfToUInt16Bits and UInt16BitsToHalf.

Implementations were based on existing signed integer variants.

Convert usages of existing APIs to unsigned variants when appropriate.

Fix #36469

The changes build on my machine, but I was unable to run the unit tests, running all of them made the test process hang for 4 hours, so I'm opening this to see what happens on the CI.

Author:	MichalPetryka
Assignees:	-
Labels:	area-System.Numerics, new-api-needs-documentation
Milestone:	-

[MichalPetryka]

I'm not quite sure what's going on with the tests failing on linux arm and musl arm & arm64, the values returned by the API seem to be way above the ushort range. Could somebody take a look?

[EgorBo]

Btw, on x64 (thus, unrelated to your issue)

        public static unsafe ulong DoubleToUInt64Bits1(double value)
        {
            Vector128<ulong> vec = Vector128.CreateScalarUnsafe(value).AsUInt64();
            return Sse2.X64.ConvertToUInt64(vec);
        }

Emits:

G_M65487_IG01:              ;; offset=0000H
       C5F877               vzeroupper 
						;; bbWeight=1    PerfScore 1.00

G_M65487_IG02:              ;; offset=0003H
       C4E1F97EC0           vmovd    rax, xmm0
						;; bbWeight=1    PerfScore 2.00

G_M65487_IG03:              ;; offset=0008H
       C3                   ret      
						;; bbWeight=1    PerfScore 1.00
; Total bytes of code: 9

@tannergooding any idea why it's movd instead of movq ? .NET 5.0 seems to be fine: https://sharplab.io/#v2:EYLgxg9gTgpgtADwGwBYA0AXEBDAzgWwB8ABAJgEYBYAKGIAYACY8gOgCUBXAOwwEt8YLAJI8ovLrl5hcAbho1iAZiakGAYQYBvedQZ79epU3JIG3XNgBmMMwBsIXAOYMAIhA7BbMACoQAqiIYqABCvBi45AAUACbunjYAbti2HDAAlDQGBtq6WXkAajBgGNDkpAAcADwc9k4AfAwJRQwAvAyFxaUVLGqw2BgwAMpgydhQfhJWMJFJKeksAIK4ATyokWlyuXkGxADsxuzcfALCouKS0iwAGuVILIO4MKTXqD0OTVAYvitBKDNFG0y2wAvjRgUA==

[EgorBo]

cc @sandreenko (#47843)

[tannergooding]

any idea why it's movd instead of movq

Looks to be a disassembly issue, its emitted an 8-byte move

[stephentoub]

Why change this one?

[MichalPetryka]

Changed it so that both places in Variant use the same pair of APIs, the unsigned ones.

[tannergooding]

LGTM.

IIRC, there were a few remaining cases that didn't immediately cast, but which should also be switched to the unsigned ones. We can get them in a follow up PR, however.

[pgovind]

Nice PR, LGTM!