Sync instruction mnemonics

[rossberg]

This updates the MVP overview to match the instruction names actually used in the test suite and interpreter.

[jakobkummerow]

It's true that I originally suggested these renamings for simplification, back when it seemed like those would be the only .new instructions.

But nowadays we have subgroup consensus to officially evaluate nominal types, which requires a second set of RTT-less .new instructions. For disassembly/debugging purposes, both opcodes (e.g. 0xfb01 and 0xfb07 could be translated to the same human-readable struct.get, but since the .wat format is also used as an assembler/interchange format by current toolchains (e.g. produced by J2CL, consumed by Binaryen), it is useful to have an unambiguous distinction between the two instructions in the text format.

So at this time, to accomplish our goals, we should maintain the distinction between:

• struct.new, 0xfb07, without RTT on the stack, with static type
• struct.new_with_rtt, 0xfb01, with RTT on the stack, possibly without static type (Remove redundant typeidx on gc instructions #241)
  (And analogously for new_default and array.new*.)

In the future, we will decide whether both instructions will be kept or only one of them; in the latter case, whichever is kept can be renamed to just struct.new for simplicity.

[tlively]

Closing this because it looks like it is no longer applicable.