Fix Socket.IO connection timeout

[Xmader]

This PR fixes the following issues:

• In XMLHttpRequest, previously we throw a DOMException(InvalidAccessError) when trying to access the xhr.withCredentials getter/setter. However in socket.io-client, it tries to set xhr.withCredentials but cannot catch the exception, thus breaks the connection.

• In our clearTimeout API, previously we expects the timeoutId argument is stored as an int32 value internally (JS numbers are always float64 values in JS context but mostly they are optimized internally). However in practice, JS number is not always optimized to an int32 value even though it's integral.
  In socket.io, old timers for the connection timeout don't get cancelled properly because the timeoutId seen as invalid, so the connection is closed prematurely.

• xhr.responseType can be one of "" | "arraybuffer" | "blob" | "document" | "json" | "text". We implemented "text" and "arraybuffer", but "json" is missing, which socket.io also requires.
  "blob" or "document" are still TODOs because there's no point outside of browser environment.

---

This PR would close #232.

[philippedistributive]

Please add a description with a reference to the issue this PR closes

Please add your posts regarding the issue from the python-monkey slack to the issue ticket so we have a clear reference and full activity log

Please document why the fixes other than commit a001cca are needed as this was not discussed in the slack chat or in the issue ticket

Also do we need to add a few tests at least no ensure no future regression?

[Xmader]

Also do we need to add a few tests at least no ensure no future regression?

Would add standardized tests for XMLHttpRequest spec compliance in later PRs.

[philippedistributive]

Let's add that as a TODO new issue

-> DONE #239

[philippedistributive]

@Xmader please see how I documented the ticket for next time around

[philippedistributive]

Let's add the code so that the TODO tests described on line 40 and 41 of test_event_loop.py pass to this PR.

See also https://developer.mozilla.org/en-US/docs/Web/API/setTimeout#non-number_delay_values_are_silently_coerced_into_numbers

for reference

Let's indeed make sure to have a full suite of tests for setTimeout/clearTimeOut while we're at it

[Xmader]

Let's add the code so that the TODO tests described on line 40 and 41 of test_event_loop.py pass to this PR.

FYI They were implemented in

PythonMonkey/python/pythonmonkey/builtin_modules/timers.js

Lines 39 to 41 in b1c3b6b

	delayMs = Number(delayMs) || 0; // ensure the `delayMs` is a `number`, explicitly do type coercion
	if (delayMs < 0)
	delayMs = 0; // as spec-ed

OK then we just needed to clean up the TODOs :-)

[philippedistributive]

Final question: why couldn't we use the XMLHttpRequest from dcp-xhr.js in dcp instead of coming up with our own?

[Xmader]

why couldn't we use the XMLHttpRequest from dcp-xhr.js in dcp

@philippedistributive Are you talking about https://gitlab.com/Distributed-Compute-Protocol/dcp/-/blob/develop/src/common/dcp-xhr.js?
It requires an external nodejs-only package xmlhttprequest-ssl, and only "patches" the XMLHttpRequest API the package provides.

coming up with our own

SpiderMonkey engine itself doesn't provide any network abilities. We need to implement the XMLHttpRequest API as well as all the network requesting stuff from ground up.

You can read https://github.com/Distributive-Network/PythonMonkey/blob/b1c3b6b/python/pythonmonkey/builtin_modules/XMLHttpRequest-internal.py

[philippedistributive]

Thanks it was a bit of a superficial, non-researched little inquiry indeed