Update the emitter to support 2-byte VEX prefix

[tannergooding]

Currently, the emitter only supports the 3-byte VEX prefix (outside a few specially handled instructions).
For normal floating-point arithmetic (or most 128-bit hardware intrinsics), this incurs an extra 1-byte per instruction.

It may be beneficial (for codegen size, and potentially throughput) to update the emitter to also support the 2-byte VEX prefix.

[tannergooding]

CC. @CarolEidt, @fiigii

[tannergooding]

This is likely easier to do now that dotnet/coreclr#19644 and dotnet/coreclr#19708 have been merged, as we can use a flag to specify that the instruction supports the 2-byte VEX encoding.