new genAst as replacement for quote do

[timotheecour]

continuation of #11722, see previous discussion in that PR. genAst subsumes quote do.

refs #11722 (comment)

closes these issues by providing a suitable replacement:

• closes Macro: bool is converted to int inside quote do #7375
• closes Strformat: undeclared identifier in quote do section. #8220 via getAstOpt({kDirtyTemplate}) see unittests
• closes Macro + Infix + quote do - compilation error #7589 (EDIT: was closed in meantime but this provides a better fix)
• closes Macros: quasiquoting + dot or bracket syntax: dot and brackets are ignored #7726
• closes quote do captures non literal local variables #9607
• closes quote do generates wrong type for set[char] #11986 (already closed but bug remains: when forgetting to use newLit, you were getting a crash)
• fixes bug 8 from [superseded] Adding support for user-defined number suffixes #17020 (comment) by using genAst instead of quote do
• plus other issues, eg recursive use, ambiguity with backticks (eg operators that need backticks don't work with quote)

for #7889, see unittests showing this is fixed (for both quote do and genAst) by using regular call syntax instead of method call syntax; can be therefore closed as dup of #7085

• closes New quoteAst RFCs#122
• supersedes [experiment] implementation for quote ast proposal #10446 and New quote ast 2 #11823

future work

• consider whether fusion/astdsl.buildAst (https://nim-lang.github.io/fusion/src/fusion/astdsl.html) should be merged inside std/genasts or in a different stdlib module; note that the use cases are different and both are useful separately (refs new genAst as replacement for quote do #17426 (comment))
• consider deprecating quote do; note that quote do can be backported to 1.0 so that code can migrate to genAst and be forward and backward compatible (it's a pure library solution).

[timotheecour]

PTAL

[timotheecour]

@Araq anything else needed for this PR?

[Araq]

Documentation. It's barely comprehensible. Take this example:

 macro fun(a: bool, b: static bool): untyped =
      let c = false # doesn't override parameter `c`
      var d = 11 # var => gensym'd
      proc localFun(): auto = 12 # proc => inject'd
      genAst(a, b, c = true):
        # echo d # not captured => gives `var not init`
        (a, b, c, localFun())
    assert fun(true, false) == (true, false, true, 12)

This is not an example how to use genAst, it tries to tell me something else, like what kind of bugs have been avoided in genAst's design. Or something else, I am not too sure what it tries to tell me.

You could also write a guest post about this new module, I can imagine lots of design effort went into it avoiding "quote do"s limitations and bugs.

[timotheecour]

PTAL, I've improved the docs and piggy backed on quote do's own example of simplified unittest.check, but made it better here by showing the lhs+rhs on error. It's intentionally simplified (eg no care taken about multiple evaluation issues or deferred failure, which are out of scope). The resulting code is compact and readable.

I also kept the other example but with more explanations

You could also write a guest post about this new module,

why not, maybe some time after this PR is merged and some initial usage feedback

[ringabout]

Cool!