WIP: Don't use stacks for simple control flow

[Keno]

Right now Zygote inserts stacks whenever it needs to use an ssa value
not defined in the first basic block. This is of course unnecessary.
The condition for needing stacks is that the basic block that defines
it is self-reachable (i.e. in a loop). Otherwise, we can simply insert
phi nodes to thread the desired SSA value through to the exit block
(we don't need to do anything in the adjoint, since the reversal of
the CFG ensures dominance). Removing stacks allows for both more
efficient code generation and enables higher order auto-diff (since
we use control flow in Zygote, but can't handle differentiating code
that contains stacks). The headline example is something like the following:

function foo(b, x)
    if b
        sin(x)
    else
        cos(x)
    end
end

Then looking at @code_typed derivative(x->foo(true, x), 1.0), we get:

Before:

CodeInfo(
1 ── %1  = $(Expr(:foreigncall, :(:jl_alloc_array_1d), Array{Int8,1}, svec(Any, Int64), :(:ccall), 2, Array{Int8,1}, 0, 0))::Array{Int8,1}
│    %2  = $(Expr(:foreigncall, :(:jl_alloc_array_1d), Array{Any,1}, svec(Any, Int64), :(:ccall), 2, Array{Any,1}, 0, 0))::Array{Any,1}
│    %3  = $(Expr(:foreigncall, :(:jl_alloc_array_1d), Array{Any,1}, svec(Any, Int64), :(:ccall), 2, Array{Any,1}, 0, 0))::Array{Any,1}
│    %4  = Base.sin::typeof(sin)
│          invoke %4(_3::Float64)::Float64
│    %6  = %new(##334#335{Float64}, x)::##334#335{Float64}
│    %7  = %new(##758#back#336{##334#335{Float64}}, %6)::##758#back#336{##334#335{Float64}}
[snip]
23 ─ %52 = invoke %47(1::Int8)::Tuple{Nothing,Nothing,Any}
│    %53 = Base.getfield(%52, 3, true)::Any
└───       goto #24
24 ─       return %53
) => Any

After (assuming some slight improvements to inference precision, that either
need to be enabled upstream or worked around here):

CodeInfo(
1 ─ %1 = Base.sin::typeof(sin)
│        invoke %1(_3::Float64)::Float64
│   %3 = Core.Intrinsics.not_int(true)::Bool
└──      goto #3 if not %3
2 ─      invoke Zygote.notnothing(nothing::Nothing)::Union{}
└──      $(Expr(:unreachable))::Union{}
3 ┄ %7 = invoke Zygote.cos(_3::Float64)::Float64
│   %8 = Base.mul_float(1.0, %7)::Float64
└──      goto #4
4 ─      goto #5
5 ─      goto #6
6 ─      goto #7
7 ─      return %8
) => Float64

Which is essentially perfect (there's a bit of junk left over, but LLVM
can take care of that. The only thing that doesn't get removed is the
useless invocation of sin, but that's a separate and known issue).

[ToucheSir]

With the benefit of a couple years of hindsight, would something like this still be worth pursuing? nested reverse AD and performance of type unstable backwards passes have gained more prominence as issues recently, and my rudimentary understanding is that this PR could address a chunk of both?