fix(accept): schecule fuzzy.access using uv.new_work

[stefanboca]

No description provided.

[Saghen]

Small nit, i think ideally the uv.new_work would be inside the access function so consumers don't have to worry about it

[stefanboca]

Good idea, I'll make the change.

This is actually not working currently, because tables (item) can't be passed as thread args. I didn't realize this at first because everything was running in a context where errors were silent. Maybe the item needs to be vim.mpack serialized/deserialized? LuaJIT also has string.buffer.{encode,decode}

[Saghen]

You might want to use the built in message pack API to serialize it before sending

[stefanboca]

Just did a quick benchmark: for a roundtrip encode/decode on a typical item, vim.mpack takes ~0.02ms, while string.buffer takes ~0.002ms. The cost is so small that IMO it's not worth dynamically detecting if string.buffer is available.

[Saghen]

Looks great, thank you!