fix: merged schemas cache key

[DamienVicet]

• run npm run test and npm run benchmark
• tests and/or benchmarks are included
• documentation is changed or added
• commit message and code follows the Developer's Certification of Origin
  and the Code of conduct

Hello, this PR fixes #724.

When diving in the bug, I found that the context.mergedSchemasIds Map, added by @ivan-tymoshenko, is using objects as keys. This caused the schema to infinitely try to build and results into Maximum call stack size exceeded error.
So I replaced the key by a hash of the object, then I got another problem : anyOf inside allOf test failed because, to my understanding, we try to cache different merged schemas using the same key in context.mergedSchemasIds (like merged schemas from buildOneOf and from buildAllOf functions).

I am not sure of the solution. Thank you for your review !

[gurgunday]

This should work, i don't know about the performance of that package, but this could be a lot faster way of setting unique identifiers:

const idMap = new WeakMap();
const getId = (obj) => {
  if (!idMap.has(obj)) {
    idMap.set(obj, Symbol('id'));
  }
  return idMap.get(obj);
};

const id = getId(object);

[gurgunday]

lgtm

[mcollina]

lgtm

I'm a bit scared of this as it would likely increase our build/startup time.

[ivan-tymoshenko]

thank you for your PR. I'm rejecting it for now, because I remember I was using objects pointers as keys on purpose to compare objects when going through the refs cycles. I need a bit more time to look at this PR. I will try to review asap.

[ivan-tymoshenko]

I found the problem. I'm working on the solution. Unfortunatelly merging schemas with recurcive references is the most complecated part of this library and take tons of time every time when we need to fix the bug.