Remove a closure allocation from WinHttpResponseStream

[jamesqo]

Pass in the current WinHttpResponseStream as a state parameter to the continuation, so we avoid a closure. Also unindented the code.

cc @davidsh @stephentoub

---

edit: Just updated it to pass in the method parameters as state, as well.

[jamesqo]

Hold on, don't merge this just yet; it looks like a closure for token is also being implicitly allocated here. Will patch in a moment.

[jamesqo]

test innerloop osx debug build and test

[jamesqo]

Maybe in the future TcsQueryDataAvailable could be changed to a subclass of TaskCompletionSource holding the relevant data to avoid the tuple allocation (similar to what was done in #7788), but for now I'm leaving this PR as-is.

[davidsh]

@jamesqo I am thinking about doing some potential refactoring of this class to address some cancellation issues when reading response streams. I want to hold this PR for now until that issue is worked out. Thanks.

[jamesqo]

@davidsh Sure, no problem.

[Petermarcu]

@davidsh , are you still working on the refactorings that you mentioned? What are the next steps on this PR?

[davidsh]

@jamesqo Thank you for your work on this PR. We are still involved in reworking parts of WinHttpResponseStream especially dealing with existing memory leaks etc. So, it would be best to close out this PR at this time.