nimgrep: change 2 iterators to closure

[a-mr]

This simple change allows to mostly eliminate nimgrep's compilation time increase (29 sec -> 6.5 sec) and code bloat (760 KB -> 540 KB) introduced in #15612.

#15924 is avoided by not destructuring tuple in the iterators' headers.

cc @timotheecour

[timotheecour]

• any chance you could show a minimized example (ideally not involving any imports) showing code that, under a similar transformation, undergoes such a drastic compilation time decrease? that will help to track if codegen does something silly
  I'm guessing this has to do with Tuple destructuring is broken with closure iterators #15924 as you mentioned

• my experience as I worked on add filewalks fusion#32 showed that closure iterators were still a bit slower than inline iterators (even in that use case), but this still seems like an acceptable tradeoff while waiting for add filewalks fusion#32 to be merged in fusion so it can be reused in nimgrep