[vm/SIMD] Vectorize ASCII decoding. #64337

I've been running some experiments on vectorizing ascii.decode through our current (and future) typed_data SIMD API surface (#64170).

The current bottleneck is the scalar loop and once we replace it with processing 16 bytes in parallel using a Uint8x16 API and a few minor optimizations outside typed_data we will be able to roughly achieve the following numbers:

We are blocked on:

• a Uint8x16 type
• Uint8x16.{signMask/anyTrue}
• Uint8x16.|
• Internal ISA flags (RISC-V is not vectorized yet, for example, and arm32 needs a separate investigation, but it should perform similar)

We've already successfully landed those members on Int32x4, so the only real blocker is the Uint8x16 type.