try fix nested dirty template

[Graveflo]

#23889

[Graveflo]

Passed CI, but of course, this solution reeks of invalid code acceptance. Seeing as how its much more annoying to make failure test cases and that failures should not be expected in the nimble packages the question is: what did I break? I dunno.

If / when this is fixed fixed, judging from the effects of this change, these additional items could be closed: #21376 #5975 #21249 #21376

[Graveflo]

well this compiles now:

proc p[T](x: T) =
  echo something

var something = 5

p(5)

clearly this change defers bindings for too long. I think it might get weird if the inner template is evaluated earlier and same if the proc's syms are bound before processing the outer template's call site. Maybe the flag in lookup is still the way to go, but only under certain conditions?

maybe that condition is just a check to see if the processing call stack is inside a template or macro, regardless of how nested it is, because in reference to above, this does make sense:

template t() =
  proc p[T](x: T) =
    echo something

var something = 5
t()
p(5)

also, no redefinition error in cases like this:

template inner(i: int) {.dirty.} =
  let thing = 1

template outer() =
  proc p[T](x: T) =
    inner(5)
    var thing = 5
    echo thing

outer()
p(0)