I totally forgot to add wake()s in multiple spots.

This commit does not change any of the code.

A subsequent commit will take advantage of this to apply the exact same tests when an EventPort is in use.

For whatever reason, the default termination handler on my machine is no longer printing the exception's `what()` string. It just aborts. That makes debugging hard.

This also means that we can now use `noexcept` in unit tests as a way to make uncaught exceptions abort the process _without_ unwinding, which is especially useful in tests that create threads since they often deadlock during unwind waiting for the thread to finish.

The new `Executor` class provides an interface to run a function on some other thread's EventLoop.

`kj::getCurrentThreadExecutor()` gets a reference to this thread's executor, which can then be exposed to other threads.

When a thread requests execution of a function on another thread, the return value is returned to the requesting thread. The requesting thread may choose to wait synchronously for this return value or -- if the requesting thread has an event loop of its own -- it can get Promise for the eventual result.

Meanwhile, orthogonally, the function can either return a raw result or return a Promise; in the latter case, the Promise is resolved to completion in the executor thread and the final result is sent back to the requesting thread.

This is better than `MutexGuarded<T>::when()` in cases where the caller already has a lock -- `when()` would require unlocking, then immediately locking again (to check the predicate).

Possibly `when()` should be removed in favor of `wait()`.

This will be important in subsequent commits adding cross-thread events to EventLoop. We'll only want wait() to throw if we know no cross-thread events will ever come, but the EventPort won't know that, only the EventLoop will.

Add basic system clock APIs.

Fix filesystem API: Empty mmaps should succeed.

I had a snarky comment in the Windows code being annoyed that mapping an empty file was documented to fail... but it turns out zero-length mappings fail on Linux, too. Also my work-around on Windows didn't work (but was never tested).

This PR fixes both.

This was prompted by @KubaO observing that `capnp compile` on an empty source file failed with a cryptic mmap error.

Closes #854, @KubaO's attempt at fixing this, since the correct fix is in the filesystem API, not in the compiler.

Fixes #822

Fixes #840

Doc fix: Expand the short description of the list pointer.

I didn't bother for posix condvars since they're kind of explicitly tied to posix clocks and it would have saved like two lines of code. But for Windows it's nice to stop using QueryPerformanceCounter() directly with the complicated math.

Add per-object streaming flow control

Implement MutexGuarded::when() (i.e. condvars) on all platforms.

Note that at present I think the only way they would have happened not under-lock is if one of the pthread or syscalls failed, which should never happen. Exceptions thrown by the predicate were already always rethrown under lock. But it doesn't hurt to be safe.

Fix sendStream() failing if it can't complete immediately.

It helps to list, in one place, both ways the list size is conveyed.

And it turns out that the Windows implementation was returning too early due to rounding error. Fixed.

It turns out GetTickCount64() is only precise to the nearest timeslice, which can be up to 16ms. The imprecision caused test failures.

I think I imagined once upon a time that this would be a convenient way to deal with external interfaces that like to return nullable pointers. However, in practice it is used nowhere in KJ or Cap'n Proto, and it recently hid a bug in my code where I had assigned a `Maybe<T>` from an `Own<T>`. We can introduce a `fromNullablePointer()` helper or something if that turns out to be useful.