byteOffset is reset to 0 for detached array buffers

[nickva]

Tried to update to the latest test262 commit in #408 and noticed a failure in the new byteOffset test: byteoffset-with-detached-buffer

In assert.sameValue(byteOffset, 2 * TA.BYTES_PER_ELEMENT); byteOffset is 8 and not 16 as the test expects.

In QuickJS we zero the byteOffset if the buffer is detached, but oddly enough, the spec doesn't call for that: [13. Let srcByteOffset be O.[[ByteOffset]]](https://tc39.es/ecma262/#sec-%typedarray%.prototype.subarray). A bit higher in If [IsTypedArrayOutOfBounds](https://tc39.es/ecma262/#sec-istypedarrayoutofbounds)(srcRecord) is true we notice the buffer is detached but that just sets the srcLength = 0 and doesn't throw an error.

[nickva]

Gave it a try here

[nickva]

It turns out I was wrong reading the spec. I started a discussion in https://es.discourse.group/t/typedarray-subarray-byteoffset-with-detached-buffers/2381 to clarify what the behavior should be. It seems it's not that we should be able to read the byteOffset of the detached buffer (we shouldn't) but that we should get the byteOffset before we get the start and the end (finish in quickjs.c) args.

The test262 test the buffer is not detached at the start of the function but only after accessing end since it has an valueOf() converter and that's where it actually detached the buffer.

I tried to fix it in #422. Sorry for confusion.