fix #14846; add macros.extractDocCommentsAndRunnables

[timotheecour]

• fix Documentation missing on devel asyncftpclient #14846 and added a regression test because this isn't the 1st time that async docgen is broken (see doc comment for async proc is empty #9816)
• the doc comments for asyncftpclient disappeared after fix #8871 runnableExamples now preserves source code comments, litterals, and all formatting; other bug fix #14439; the problem however is not fix #8871 runnableExamples now preserves source code comments, litterals, and all formatting; other bug fix #14439 but it's caused by async which inserted doc comments after a block that added template definitions, resulting in something like:

proc main*() =
  template gun()=discard
  ## in main
  discard

before #14439, the doc comment would be docgen'd even though the doc comment is at wrong place (after some initial non-doc comment/non-runnableExamples statement)
after #14439, in main correctly doesn't get docgen'd

this PR fixes async template by introducing extractDocCommentsAndRunnables which extracts all top-level runnableExamples and doc comments

after this PR

after this I have another PR that removes the awaitDefinition block which redefines await templates in each each async proc, and instead bindSym's a await template defined once, which simplifies the generated code

[juancarlospaco]

Why not 2 extractComments() and extractRunnables(), shorter name and differentiates both instead of mixing them.

[timotheecour]

because of interleaving; having to procs would not preserve interleaving eg:

proc fn*=
  ## c1
  runnableExamples:
    discard
  runnableExamples:
    discard
  ## c2
  runnableExamples:
    discard
  ## c3
  echo "start of actual code"

(interleaving is supported since #14441; before that it was buggy as it dropped the doc comments after the 1st one)

[Clyybber]

IMO docgen should simply support this pattern instead (the doc comment appearing after template definitions/other code).

[timotheecour]

we can debate that in nim-lang/RFCs#231 but until then this PR fixes docgen for async procs; one potential problem is doc comments appearing inside proc body but not intended to be docgen'd; another one is doc comments in templates called from a proc.

[timotheecour]

the failure is potentially caused by this PR, not sure yet

2020-06-30T18:09:23.7296506Z FAIL: tests/async/testmanyasyncevents.nim C
2020-06-30T18:09:23.7297206Z Test "tests\async\testmanyasyncevents.nim" in category "async"
2020-06-30T18:09:23.7298204Z Failure: reOutputsDiffer
2020-06-30T18:09:23.7298736Z Expected:
2020-06-30T18:09:23.7299265Z hasPendingOperations: false
2020-06-30T18:09:23.7299819Z triggerCount: 100
2020-06-30T18:09:23.7300206Z 
2020-06-30T18:09:23.7300642Z Gotten:
2020-06-30T18:09:23.7301187Z hasPendingOperations: true
2020-06-30T18:09:23.7301763Z triggerCount: 68
2020-06-30T18:09:23.7302159Z