StdInitializerListOp next steps

[HerrCai0907]

I was giving this more thought, and I think we should take a step back:

• We should revert the PR that introduced cir.std.initializer_list.
• Introduce a new PR where cir.std.initializer_list is just a thin layer over the initialization work, for example:

%0 = cir.alloca INITLIST_TYPE, !cir.ptr<INITLIST_TYPE>
cir.std.initializer_list {
  %1 = cir.const #cir.int<1>
  ... // all init work goes here
}

My rationale is that we don't yet have a pass taking advantage of this and we should probably use that to guide how we really want to design this, anything before that will be more or less a guess game. Once we have usage going through it and we start writing a pass that wants to look at the initialization data, then we refine. Sorry for the churn, WDYT?

Originally posted by @bcardosolopes in #1184 (comment)