Tracking instructions with unassigned opcodes

[tlively]

We don't want to do another full renumbering, but we also want to choose reasonable opcodes for new instructions, so we decided here and in the 1/8/21 meeting that beginning with #376, the final opcodes of all newly merged instructions will be decided after we finalize the instruction set. The opcodes of instructions merged earlier than #376 will not change.

Here is a (hopefully up-to-date) list of instructions whose opcodes will need to be finalized:

• Extended multiplications (#376)
• Load/store lane instructions (#350)
• i64x2 widening instructions (#290)
• Saturating rounding Q-format multiplication (#365)
• i64x2.bitmask (#368)

[Maratyszcza]

A couple more:

• v128.any_true instruction to replace i8x16.any_true/i16x8.any_true/i32x4.any_true/i64x2.any_true (#416)
• Potentially floating-point rounding instructions, as their opcodes overlap with 64-bit min/max reserved opcodes

[ngzhian]

Closing, this is done, and superseded by #452