[Web] Fix Webkit leak caused by the position reporting audio worklets

[adamscott]

This PR reintroduces the pooling of GodotPositionReportingProcessor worklets. This is because Webkit doesn't seem to delete processors that return false in process(), even if the specs says to do so. That made the memory explode and the Webkit engine reloading the games.

This PR also removes the postMessage limit on audio position worklet. The tests made by @PizzaLovers007 seem to indicate that the partial reporting didn't achieve it's goals and was basically useless.

Here's my iPad Pro M3 running a tweaked version of the #107390 MRP for more than 18 minutes straight.

At the 18 minutes mark, nothing seems to indicate that the MRP was taxing for my iPad, memory-wise nor processing-wise.

MRP: audio-test.zip

Fixes #107390.
Fixes #105397.
Supersedes #105541. (added co-authorship)
Reverts the spirit of #102715.
Doesn't seem to create a regression for #102505.

[adamscott]

cc. @rburing @mooflu @PizzaLovers007

[mellowminx]

Thanks so much for this fix! 🥳

[Zireael07]

So we're effectively going back and forth on those audio worklets :/

[akien-mga]

Thanks!

[adamscott]

So we're effectively going back and forth on those audio worklets :/

I think this one is for good.

[mellowminx]

I excitedly downloaded Godot 4.5 beta 2 to test this fix, thank you so much! Working for me, tested for at least 30 mins each on iOS Chrome and iOS Safari. Test export is here: https://mellowminx.itch.io/audio-test-godot-45-beta2 (password: godot)

[Frozenfire92]

Does this change require a recompile? I'm wondering if its possible to fix for 4.4.1

I tried copying the changes here in my two javascript files on my 4.4.1 export and it got further than before (and audio played), it still crashed/reloaded after about 30 seconds