[spec] Support i8 and i16?

[achun]

previous discussion: #1049 #85 #804

related proposal: sign-extension-ops simd

I will not repeat the comments of previous here.

In fact, the instructions are always supporting both types.
And new proposals have to consider and support them.

in fact, signed integers have always existed and are supported, in all numerical instructions. However, they are not in the number types: s8, s16, s32, s64

The instructions:

• i64.load8_s means s8.load and s64.extend_s8
• i32.extend8_s means s32.extend_s8
• i8x16.splat means v128.splat_i8x16

We know that the number of 1-byte-opcode is limited. It is not necessary to allocate instructions like this i16.add. Assigning a prefix byte to them is worth considering.

They exist in the instruction but are not named, very strange.