WinHttpHandler: fix double -> uint conversions for ARM64

[antonfirsov]

According to the Explicit numeric conversions section of the C# specification, the unchecked case of a double -> uint cast goes as following:

• If the value of the operand is NaN or infinite, the result of the conversion is an unspecified value of the destination type.
• Otherwise, the source operand is rounded towards zero to the nearest integral value. If this integral value is within the range of the destination type then this value is the result of the conversion.
• Otherwise, the result of the conversion is an unspecified value of the destination type.

Looks like we do not handle the Timeout.InfiniteTimeSpan.TotalMilliseconds == -1.0 case correctly for ReceiveDataTimeout and TcpKeepAlive* , the latter resulting in test failures on ARM64 machines.

Fixes #45249

Note:
The failing Libraries Test Run release coreclr windows arm64 Release job does not get executed on PR-s, only on refs/heads/master, which means there is no way for me to validate the change before merging it. (Unless someone can suggest a way to access an ARM64 windows machine.)

/cc @EgorBo @tannergooding if interested.

[ghost]

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

Issue Details

---

According to the Explicit numeric conversions section of the C# specification, the unchecked case of a double -> uint cast goes as following:

• If the value of the operand is NaN or infinite, the result of the conversion is an unspecified value of the destination type.
• Otherwise, the source operand is rounded towards zero to the nearest integral value. If this integral value is within the range of the destination type then this value is the result of the conversion.
• Otherwise, the result of the conversion is an unspecified value of the destination type.

Looks like we do not handle the Timeout.InfiniteTimeSpan.TotalMilliseconds == -1.0 case correctly for ReceiveDataTimeout and TcpKeepAlive* , the latter resulting in test failures on ARM64 machines.

Fixes #45249

Note:
The failing Libraries Test Run release coreclr windows arm64 Release job does not get executed on PR-s, only on refs/heads/master, which means there is no way for me to validate the change before merging it. (Unless someone can suggest a way to access an ARM64 windows machine.)

/cc @EgorBo @tannergooding if interested.

Author:	antonfirsov
Assignees:	-
Labels:	area-System.Net.Http
Milestone:	-

[CarnaViire]

A tiny nit, otherwise LGTM

[antonfirsov]

Test failures are unrelated.

[safern]

Failures are known and unrelated. Merging to unblock CI.

[ViktorHofer]

Failures are:

• SafeHandle use-after-dispose in FileSystemWatcher on OSX #30056
• FrameworkHiveSelection_GlobalHiveWithBetterMatch failed in CI #45510

(was looking at it while Santi merged the PR)

[ViktorHofer]

Test failures are unrelated.

@antonfirsov even if test failures are unrelated, in future please search for existing issues or create them and link to the affected build so that the failures are refcounted.