toSeq should use copyMem when possible

[timotheecour]

unless we can guarantee that existing code in toSeq produces the most efficient code in release, we should make toSeq should use copyMem when possible

example

import std/sequtils

type Foo = object
  f0: seq[int]

proc fn(a1: openArray[int])=
  var a2 = [1,2]
  var a3 = @[1,2]
  # var f = Foo(f0: a1) # Error: type mismatch:
  # var f = Foo(f0: a2) # ditto
  # var s = a1 # ditto
  # var f = Foo(f0: a3) # ook
  var s = a1.toSeq # works but via a for-loop (currently); toSeq should use copyMem when possible

fn([1,2])

notes

ideally, copyMem could have a safer wrapper that can be reused (with no performance penalty)
(maybe in strbasics? or rename this module since lots of low-level procs are similar for seq/string?)

links

• refs followup now that https://github.com/nim-lang/Nim/issues/15511 was fixed by timotheecour · Pull Request #103 · nitely/nim-regex