RFR(s): 8218041: Assorted wrong/missing includes

[Robbin Ehn]

> I agree with this move. try_cancel is not very performance-sensitive, and it is called when other,
> more heavy-weight stuff is happening. I marked the bug with "gc-shenandoah", so we get it in our
> backporting queue. Shenandoah still builds fine with/without PCH, and passes tests after this change.

Great, and it still can be inlined since it's in the same compilation unit and 
you do not take the address of it.

> 
> Other changes look good too.

Thanks, v02 in a sec or two!

/Robbin

> 
> -Aleksey
>