[mypyc] Support top level function ops via CallC

[TH3CHARLie]

relates mypyc/mypyc#709

This PR supports top-level function ops via recently added CallC IR. To demonstrate the idea, it transform to_list op from PrimitiveOp to CallC. It also refines CallC with arguments coercing and support of steals.

[TH3CHARLie]

Matching func_ops(where to_list op originally resides) happens in two places: ll_builder/builtin_call and builder/call_refexpr_with_args. I changed them accordingly, but I also observe that for to_list case(at least for simple cases I tested), only the latter one is doing the work.

[JukkaL]

Looks great!

[JukkaL]

Let me know when you are ready with the PR (it now says WIP).

[TH3CHARLie]

Let me know when you are ready with the PR (it now says WIP).

Yes, I'd say it's WIP because I intend to include list_set_item_op as well and there are some other things that may need to discuss.

[TH3CHARLie]

I think it's now ready for review(assuming all tests passed).

Updates(57a1b91): Well, many tests failed. I observe and summarize that are two kinds of failing tests. The first one is IR tests, which can be fixed quickly. The second kind is about traceback, changing list_set_item op somehow produces much less traceback info.

updates(983d635): I revert list_set_item_op as I think it is too intrusive to test suite. Supporting it requires changing a lot of tests and make the PR pretty huge. And figuring out why changing it from PrimitiveOp to CallC causes different traceback behavior needs time and discussion.

[TH3CHARLie]

I delete a blank line here and I find that one/two blank line(s) is used interchangeably. I am surprised that flake8 is OK with this

[JukkaL]

Left some comments about coercing the result type.

[JukkaL]

Thanks for the updates! Looks great.

[TH3CHARLie]

Thanks for the review!