Add support for Linker defined host functions

[Robbepop]

Closes #663.

cc @flavio

Todo

• Add tests for Linker::func_new and Linker::func_wrap.

[paritytech-cicd-pr]

BENCHMARKS

	NATIVE			WASMTIME
BENCHMARK	MASTER	PR	DIFF	MASTER	PR	DIFF	WASMTIME OVERHEAD
execute/ bare_call_0	1.51ms	1.51ms	⚪ 0.07%	1.46ms	1.37ms	🟢 -5.95%	🟢 -9%
execute/ bare_call_0/typed	1.11ms	1.12ms	⚪ 0.67%	954.89µs	956.23µs	⚪ 0.00%	🟢 -15%
execute/ bare_call_1	1.55ms	1.56ms	🔴 0.61%	1.48ms	1.45ms	🔴 -2.03%	🟢 -7%
execute/ bare_call_16	2.45ms	2.47ms	🔴 0.37%	3.31ms	3.44ms	🔴 3.98%	🟢 40%
execute/ bare_call_16/typed	1.75ms	1.82ms	🔴 3.85%	1.60ms	1.63ms	🔴 1.83%	🟢 -10%
execute/ bare_call_1/typed	1.22ms	1.23ms	⚪ 0.50%	1.31ms	1.24ms	🟢 -5.46%	🟢 0%
execute/ bare_call_4	1.72ms	1.74ms	🔴 1.41%	1.78ms	1.77ms	🔴 -0.22%	🟢 2%
execute/ bare_call_4/typed	1.27ms	1.31ms	🔴 3.36%	1.17ms	1.22ms	🔴 3.77%	🟢 -7%
execute/ br_table	1.30ms	1.31ms	⚪ 0.74%	1.41ms	1.39ms	🟢 -1.25%	🟢 6%
execute/ count_until	718.03µs	680.67µs	🟢 -5.12%	1.10ms	989.99µs	🟢 -9.90%	🟢 45%
execute/ factorial_iterative	368.86µs	332.49µs	🟢 -9.68%	485.66µs	433.91µs	🟢 -10.51%	🟢 31%
execute/ factorial_recursive	595.35µs	588.22µs	⚪ -1.03%	900.61µs	883.16µs	🟢 -1.72%	🟢 50%
execute/ fibonacci_iter	1.47ms	1.47ms	⚪ -0.47%	2.06ms	2.06ms	⚪ -0.11%	🟢 40%
execute/ fibonacci_rec	4.83ms	4.77ms	⚪ -1.19%	8.35ms	7.23ms	🟢 -13.44%	🟡 52%
execute/ fibonacci_tail	1.22ms	1.18ms	🟢 -3.83%	1.77ms	1.56ms	🟢 -12.41%	🟢 32%
execute/ global_bump	798.70µs	786.86µs	⚪ -1.13%	1.56ms	1.32ms	🟢 -15.23%	🟡 68%
execute/ global_const	812.12µs	718.80µs	🟢 -11.44%	1.39ms	1.17ms	🟢 -15.55%	🟡 63%
execute/ host_calls	35.64µs	35.92µs	⚪ 0.72%	47.01µs	48.86µs	🔴 4.04%	🟢 36%
execute/ memory_fill	1.21ms	1.20ms	⚪ -1.33%	1.78ms	1.74ms	🟢 -2.61%	🟢 45%
execute/ memory_sum	1.19ms	1.19ms	⚪ -0.06%	1.85ms	1.76ms	🟢 -4.87%	🟢 49%
execute/ memory_vec_add	2.38ms	2.40ms	⚪ 0.62%	4.66ms	3.88ms	🟢 -16.62%	🟡 62%
execute/ recursive_is_even	1.04ms	1.05ms	⚪ 0.65%	1.67ms	1.75ms	🔴 5.01%	🟡 67%
execute/ recursive_ok	123.45µs	120.86µs	🟢 -2.28%	203.38µs	204.13µs	⚪ 0.11%	🟡 69%
execute/ recursive_scan	161.16µs	157.27µs	⚪ -2.00%	261.91µs	252.97µs	🟢 -3.51%	🟡 61%
execute/ recursive_trap	12.53µs	12.34µs	🟢 -1.47%	20.92µs	22.50µs	🔴 7.44%	🟡 82%
execute/ regex_redux	500.23µs	498.97µs	⚪ -0.36%	898.08µs	841.09µs	🟢 -6.36%	🟡 69%
execute/ rev_complement	464.25µs	462.10µs	⚪ 0.32%	778.03µs	723.57µs	🟢 -7.03%	🟡 57%
execute/ tiny_keccak	325.60µs	324.37µs	⚪ -0.13%	575.63µs	546.44µs	🟢 -5.15%	🟡 68%
execute/ trunc_f2i	787.73µs	779.86µs	⚪ -0.97%	1.38ms	1.37ms	⚪ -1.01%	🟡 75%
instantiate/ wasm_kernel	56.32µs	54.06µs	🟢 -2.81%	58.80µs	55.35µs	🟢 -6.58%	🟢 2%
translate/ erc1155	234.79µs	234.79µs	⚪ -0.46%	432.97µs	440.45µs	🔴 1.75%	🟡 88%
translate/ erc20	113.75µs	114.06µs	⚪ 0.39%	210.39µs	212.22µs	⚪ 1.15%	🟡 86%
translate/ erc721	160.45µs	162.34µs	⚪ 0.79%	307.69µs	309.86µs	⚪ 0.88%	🟡 91%
translate/ spidermonkey	0.00ns	0.00ns	🔴 1.66%	0.00ns	0.00ns	🔴 1.56%	🟢 0%
translate/ wasm_kernel	4.04ms	4.10ms	🔴 1.56%	8.09ms	8.18ms	⚪ 1.09%	🟡 99%

Link to pipeline

[Robbepop]

As of now I don't know where the double digit performance regressions in some native benchmarks are coming from. I can reproduce them locally. Will need some research about this before merging this PR. It is very weird since this PR does not even remotely touch the execution engine of wasmi.

[Robbepop]

This PR in its current state is not mergeable due to the massive performance regressions on the Wasm target.

[Robbepop]

Performance regressions seems to be resolved with the most recent merge with master so we can finally merge this PR.

[flavio]

Sweet, I'll update my wit-bindgen code to make use of it :)