Add op-assign instructions to wasmi IR

[Robbepop]

Closes #767.

TODO

• Define new op-assign instructions in wasmi IR.
• Implement execution of new op-assign instructions.
• Implement translation of op-assign instructions.
• Implement optimizations from or to op-assign instructions.
  • local.set: op-assign -> op
  • local.set: op -> op-assign
  • cmp-assign+branch -> fuse
• Add tests to cover new functionality.

[paritytech-cicd-pr]

BENCHMARKS

	NATIVE			WASMTIME
BENCHMARK	MASTER	PR	DIFF	MASTER	PR	DIFF	WASMTIME OVERHEAD
execute/ bare_call_0	1.55ms	1.65ms	🔴 6.95%	1.18ms	1.16ms	🔴 -1.96%	🟢 -30%
execute/ bare_call_0/typed	1.17ms	1.17ms	⚪ -0.84%	801.20µs	928.34µs	🔴 15.87%	🟢 -20%
execute/ bare_call_1	1.62ms	1.70ms	🔴 5.19%	1.28ms	1.31ms	🔴 2.46%	🟢 -23%
execute/ bare_call_16	2.52ms	2.52ms	🔴 -0.18%	3.55ms	3.56ms	🔴 0.59%	🟢 41%
execute/ bare_call_16/typed	1.55ms	1.63ms	🔴 5.01%	1.95ms	2.00ms	🔴 2.47%	🟢 23%
execute/ bare_call_1/typed	1.27ms	1.29ms	🔴 1.80%	998.11µs	996.96µs	⚪ -0.07%	🟢 -23%
execute/ bare_call_4	1.78ms	1.78ms	🔴 0.00%	1.71ms	1.63ms	🔴 -4.75%	🟢 -8%
execute/ bare_call_4/typed	1.22ms	1.32ms	🔴 8.69%	1.05ms	1.11ms	🔴 5.64%	🟢 -16%
execute/ br_table	1.39ms	1.37ms	⚪ -1.07%	1.14ms	1.18ms	🔴 3.54%	🟢 -14%
execute/ count_until	546.17µs	545.74µs	⚪ -0.10%	1.69ms	1.68ms	⚪ -0.79%	🔴 207%
execute/ factorial_iterative	322.11µs	320.17µs	⚪ -0.61%	805.50µs	805.50µs	⚪ 0.06%	🔴 152%
execute/ factorial_recursive	490.12µs	493.30µs	⚪ 0.62%	969.61µs	969.54µs	⚪ -0.02%	🟡 97%
execute/ fibonacci_iter	1.40ms	1.40ms	⚪ 0.34%	3.81ms	3.84ms	⚪ 0.36%	🔴 174%
execute/ fibonacci_rec	4.04ms	4.00ms	⚪ -1.09%	8.59ms	8.56ms	⚪ -0.35%	🔴 114%
execute/ fibonacci_tail	859.45µs	864.14µs	⚪ 0.57%	2.20ms	2.17ms	⚪ -1.08%	🔴 151%
execute/ global_bump	746.86µs	741.03µs	⚪ -0.78%	2.19ms	2.20ms	⚪ 0.46%	🔴 197%
execute/ global_const	659.02µs	660.22µs	⚪ 0.08%	2.44ms	2.42ms	⚪ -0.60%	🔴 267%
execute/ host_calls	37.17µs	36.05µs	🟢 -3.19%	39.87µs	40.38µs	🔴 1.23%	🟢 12%
execute/ memory_fill	1.15ms	1.21ms	🔴 5.55%	3.32ms	3.30ms	⚪ -0.61%	🔴 172%
execute/ memory_sum	1.14ms	1.18ms	⚪ 2.43%	3.29ms	3.30ms	⚪ 0.26%	🔴 179%
execute/ memory_vec_add	2.34ms	2.39ms	⚪ 1.85%	7.41ms	7.39ms	⚪ -0.22%	🔴 209%
execute/ recursive_is_even	663.42µs	662.37µs	⚪ -0.14%	1.47ms	1.45ms	🟢 -1.18%	🔴 119%
execute/ recursive_ok	94.34µs	94.01µs	⚪ -0.47%	200.90µs	200.71µs	⚪ -0.07%	🔴 114%
execute/ recursive_scan	129.19µs	129.42µs	⚪ 0.38%	284.87µs	287.38µs	⚪ 0.88%	🔴 122%
execute/ recursive_trap	8.80µs	8.90µs	🔴 1.16%	20.96µs	20.94µs	⚪ -0.05%	🔴 135%
execute/ regex_redux	457.04µs	463.29µs	🔴 1.45%	1.23ms	1.24ms	⚪ 0.40%	🔴 167%
execute/ rev_complement	424.09µs	422.19µs	⚪ -0.46%	1.14ms	1.17ms	🔴 2.34%	🔴 176%
execute/ tiny_keccak	317.45µs	323.09µs	🔴 1.73%	1.09ms	1.10ms	⚪ 0.42%	🔴 240%
execute/ trunc_f2i	737.07µs	732.38µs	⚪ -0.56%	1.71ms	1.71ms	⚪ 0.09%	🔴 134%
instantiate/ wasm_kernel	54.94µs	55.79µs	🔴 3.46%	56.22µs	58.93µs	🔴 2.72%	🟢 6%
translate/ erc1155	210.40µs	208.07µs	⚪ -1.07%	385.12µs	364.78µs	🟢 -5.40%	🟡 75%
translate/ erc20	103.72µs	103.67µs	⚪ -0.09%	182.12µs	178.39µs	🟢 -2.04%	🟡 72%
translate/ erc721	145.99µs	146.34µs	⚪ 0.08%	263.76µs	257.33µs	🟢 -2.53%	🟡 76%
translate/ spidermonkey	64.01ms	64.04ms	⚪ -0.22%	0.00ns	0.00ns	🟢 -2.08%	🟢 -100%
translate/ wasm_kernel	4.17ms	4.18ms	⚪ 0.30%	6.79ms	6.61ms	🟢 -2.88%	🟡 58%

Link to pipeline

[Robbepop]

The performance improvements are extremely modest given that this PR is adding well over 100 new wasmi IR instructions. So either we are doing something wrong or this whole effort is definitely not worth it.

[Robbepop]

Closed since deemed not to be worth the efforts performance wise.