sigmatch: support optional params with last block arg(s)

[timotheecour]

this PR solves a major pain point of optional params with last block arg(s)

fix nim-lang/RFCs#405
fix #884
fix #14346

template fn(a = 1, b = 2, body) = discard
fn(1, 2): # already works
  bar
fn(a = 1): # now works
  bar

ditto with multiple block args via do:

template fn(a = 1, b = 2, body1, body2) = discard
fn(a = 1): # now works
  bar1
do:
  bar2

etc, see tests/misc/trfc405.nim

[c-blake]

Maybe one of your tests should check the >1 body passed case with a do:?

[timotheecour]

Maybe one of your tests should check the >1 body passed case with a do:?

i was planning on supporting that in future work, but i've now done it, see tests

[disruptek]

Why would I want this? To make statement lists more ambiguous?

[disruptek]

I think I simply don't understand your example. Can you explain why you added bar to the block?

[disruptek]

For that matter, can you explain why you add when true:? This is noise and I am tired of reading it.

[timotheecour]

Why would I want this

because optional params are useful and using overloads instead causes known problems, see linked issues

[Araq]

The implementation matches the RFC, the only minor gripe would be that you should extract the new search loop into a helper proc. Will comment on the RFC.

[timotheecour]

you should extract the new search loop into a helper proc. Will comment on the RFC.

done, PTAL

[timotheecour]

ping @Araq , this is a big usability improvement; (and I've replied in nim-lang/RFCs#405 (comment) to your comment in RFC)

[Clyybber]

This caused a regression reported on the forums: https://forum.nim-lang.org/t/8513
EDIT: Also reported on github #19015

[solo989]

This could work properly with varargs if the matching algorithm matched the first non default argument that hasn't already been matched instead of working backwards like it does now.

That way it wouldn't try to fill the varargs argument in the case where it should fill other non default arguments first.