Add emscripten_promising_count() for querying in-flight async invocations

[guybedford]

This adds a new emscripten_promising_count() function returning the number of async (promising) wasm invocations currently in flight, entered but with their returned promise not yet settled, under both ASYNCIFY and JSPI.

This allows runtime code to detect whether any async wasm invocations are pending, for example to verify or assert reentrancy invariants.

• For JSPI (ASYNCIFY==2), the counter is bumped in Asyncify.makeAsyncFunction around each promising invocation, decrementing when the returned promise settles. All WebAssembly.promising usage (exports, dynCall, embind, makeDynCall codegen) is now routed through makeAsyncFunction as a single chokepoint, so this is a single bump per invocation rather than per import call.
• For ASYNCIFY, the single suspendable invocation is counted from its first unwind until its final rewind completes, so the value is 0 or 1 (fiber swaps are not counted).
• Without async support the function compiles to a constant 0.

One documented asymmetry: under JSPI a promising export counts for its entire invocation, while under ASYNCIFY an invocation is only counted once it first suspends.

Includes a new test_promising_count core test run under both ASYNCIFY and JSPI, observing the count from JS during an emscripten_sleep suspension, from C after resume, and confirming it returns to 0 after the invocation settles. Docs and ChangeLog updated.

Made with AI assistance under my review

[sbc100]

I think the current state is the we only support a single suspended stack at once, either with ASYNCIFY or JSPI? At least I think it should be.

If I'm wrong and we don't currently enforce this for JSPI, then perhaps we should land the internals of this counter, along with an assertion if its ever exceeds 1?

We can then add a the public function, once we add the ASYNCIFY_RENTRANT option (until we add that we should not allow more than one suspended stack I think).

If/when we do add a public function here, perhaps it should be called something like emscripten_suspended_stack_count? Or emscripten_asyncify_active_stacks? Just shedding here.