vm: allow sharing a microtask queue across contexts

[agape1225]

Extends the existing microtaskMode option (no new option) so multiple
vm contexts can share one explicitly-drained microtask queue:

const queue = vm.createMicrotaskQueue();
const window = vm.createContext(windowSandbox, { microtaskMode: { type: 'manual', queue } });
const iframe = vm.createContext(iframeSandbox, { microtaskMode: { type: 'manual', queue } });

vm.runInContext('...', window); // not drained automatically
queue.runMicrotasks();          // drained explicitly, once, by the embedder

Today, 'afterEvaluate' gives each context its own private, auto-drained
queue — there's no way for two contexts to share one and have it drained
together, in scheduling order, as the HTML spec requires for same-agent
realms (e.g. a Window and a same-origin iframe).

Design

• microtaskMode now accepts either the existing string 'afterEvaluate'
  or an object { type: 'manual', queue }, where queue is created via
  the one new export, vm.createMicrotaskQueue(). These are two values
  of the same option, so a context is either auto-drained with a private
  queue or shared-and-manual — never both, by construction, with nothing
  to validate for that combination at runtime.
• The native binding's argument list and MakeContext() signature are
  unchanged; the manual-mode queue is passed through the same slot that
  used to only take a boolean.
• The queue handle (ContextifyMicrotaskQueue) is a plain BaseObject:
  the underlying v8::MicrotaskQueue is kept alive via std::shared_ptr,
  shared between the JS handle and every ContextifyContext using it, so
  the handle itself doesn't need cppgc tracking.
• Passing anything other than a real vm.MicrotaskQueue (a spoofed
  object, a Proxy, etc.) is rejected with ERR_INVALID_ARG_TYPE,
  checked via FunctionTemplate::HasInstance() rather than a JS-level
  instanceof.

Testing

New tests in test/parallel/test-vm-shared-microtask-queue*.js cover
the issue's Window/iframe scenario, queue-spoofing rejection, GC of the
JS handle, and module evaluation. make -j12 and the full
test/parallel/test-vm-*.js suite pass (104/104).

Refs: #65555
Refs: #65582

[codecov]

Codecov Report

❌ Patch coverage is 93.06931% with 7 lines in your changes missing coverage. Please review.
✅ Project coverage is 90.41%. Comparing base (a18fd89) to head (eb20179).
⚠️ Report is 179 commits behind head on main.

Files with missing lines	Patch %	Lines
src/node_contextify.cc	91.11%	0 Missing and 4 partials ⚠️
src/node_contextify.h	40.00%	3 Missing ⚠️

Additional details and impacted files

@@            Coverage Diff             @@
##             main   #66461      +/-   ##
==========================================
+ Coverage   90.33%   90.41%   +0.08%     
==========================================
  Files         790      791       +1     
  Lines      273834   275693    +1859     
  Branches    52361    52861     +500     
==========================================
+ Hits       247378   249280    +1902     
+ Misses      16924    16817     -107     
- Partials     9532     9596      +64     

Files with missing lines	Coverage Δ
lib/vm.js	99.35% <100.00%> (+0.06%)	⬆️
src/node_contextify.h	63.63% <40.00%> (-6.96%)	⬇️
src/node_contextify.cc	81.99% <91.11%> (+0.18%)	⬆️

... and 143 files with indirect coverage changes

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.
• 📦 JS Bundle Analysis: Save yourself from yourself by tracking and limiting bundle sizes in JS merges.