[Experimental] Implement writeIntoUint8Array for TextEncoder

[jasnell]

This is an experiment for now. Do not merge yet.

Background: When running under high memory pressure, operations like TextEncoder::encode can cause additional memory load by flattening the input strings. This is particularly true in React/NextJS SSR during the render phase. The string flattening itself can lead to additional GC thrashing that adds significant additional CPU overhead to the rendering. This PR experiments with detecting when we are in high memory load conditions and using an alternative utf8 encode that avoids the string flattening. It only kicks in under certain conditions (e.g. string isn't already flat, string is over a certain length, memory pressure is over a certain threshold, etc). It's not intended to make encoding faster, it is ended to avoid flattening and the associated GC costs.

There's an additional v8 patch included, so this PR is not expected to pass internal CI until the same patch is landed there.

[jasnell]

btw, in case it's not obvious, just had claude do the tests.

[jasnell]

Ok, given a bunch of additional testing and evaluation and review from Erik, this is not the direction we want to go.. which is pretty much what I suspected :-) .. closing this!