Missed one backing field.

Still a lot of docs work to follow-up with here.

Lots of cleanup here to make additional bits later more straightforward and less conflicting.

Same order where possible and eliminated the Browsable attribute diffrence with a shim instead.

After 50547eba, TaskSource.Create and TaskSource.CreateDenyExecSync methods are the same. There was some crud left everywhere that appeared like differences but it's really determined once in the static initializer for TaskSource based on the platform anyway.

I've fixed the tests as well - they were a bit backwards on logic/assumptions so hard to read, and have been failing for some time. Now we'd expect ContinueWithExecSync to *not* be hijacked, and everything else to be...that's how the tests are as well.

Note: the #if DEBUG change is because internals are already visible to tests...just simpler.

This doesn't resolve the issue, but adds several confirmations and more information. At the moment we're not recognizing :6380 as a slaved instance though it clearly is from the outside. There's more to dig into on recognition here. This just improves the test quite a bit, before it was throwing in the middle without much info.

This didn't prctically matter because they match with defaults...but was wrong.

This failed under unit test load due to unintended retries during periods where async continuations for completion of the test don't occur before retries kick in and change an auth exception to a socket exception (which we correctly already closed).

Still, I think we should consider storing the initial connection error seperately from others as there's a fundamental race of an error that an observer cannot see when retries overwrite the last error message.

Previously was allowing 0ms to actually get a broadcast

TODO: inherit from tests project and use a proper Skip.

Also fixes tests that go along with (failing previously due to a message change before this one).