inlining: remove ineffective handling for unmatched params

[aviatesk]

The deleted branch was added in #45062, although it had not been tested.

I tried the following diff to find cases optimized by that, but I just found the handling proved to be in vain in all cases I tried.

diff --git a/base/compiler/ssair/inlining.jl b/base/compiler/ssair/inlining.jl
index 318b21b09b..7e42a65aa4 100644
--- a/base/compiler/ssair/inlining.jl
+++ b/base/compiler/ssair/inlining.jl
@@ -1473,6 +1473,14 @@ function compute_inlining_cases(@nospecialize(info::CallInfo), flag::UInt32, sig
         handle_any_const_result!(cases,
             result, match, argtypes, info, flag, state; allow_abstract=true, allow_typevars=true)
         fully_covered = handled_all_cases = match.fully_covers
+        if length(cases) == 1 && fully_covered
+            println("first case: ", only_method)
+        elseif length(cases) == 1
+            atype = argtypes_to_type(sig.argtypes)
+            if atype isa DataType && cases[1].sig isa DataType
+                println("second case: ", only_method)
+            end
+        end
     elseif !handled_all_cases
         # if we've not seen all candidates, union split is valid only for dispatch tuples
         filter!(case::InliningCase->isdispatchtuple(case.sig), cases)

[Keno]

This code is not supposed to be dead. Here's a test case that's supposed to trigger it:

function foo(x::Union{Int, Float64}, y::T) where {T}
    Base.donotdelete(T)
end
function bar(b::Bool, @nospecialize(y))
     foo(b ? 1 : 1.0, y)
end

I think this broke when we started relying more heavily on inference to push forward info objects.

[aviatesk]

I found this fails because there is no cached source to be inlined available for foo(:Union{Int, Float64}, y::T) (in common cases) since inference union-splits, although inlining will hit this branch in such a case:

julia/base/compiler/ssair/inlining.jl

Lines 1452 to 1456 in 85d7cca

	if nsplit(info)::Int > 1
	atype = argtypes_to_type(argtypes)
	(metharg, methsp) = ccall(:jl_type_intersection_with_env, Any, (Any, Any), atype, only_method.sig)::SimpleVector
	match = MethodMatch(metharg, methsp::SimpleVector, only_method, true)
	result = nothing

[aviatesk]

I don't think it's profitable to try union-split ininling with unmatched parameters in such a case, since SROA can't eliminate them anyway.

[aviatesk]

I'm thinking of moving this PR forward first to simplify the fix for #52644. Right now, the handling we're talking about isn't really turned on. If we decide to add similar features in the future, we'll likely need some extra help from the inference side of things. When that happens, we can bring this back with some effective test cases.

[vtjnash]

Why do we set handled_all_cases if we didn't handle everything? This comment seems confusing here

[aviatesk]

revisit_idx might be effective even when there are other matches that have been handled properly. But I agree that this can be confusing. I'm working on some more clean up.