Improve nothrow analysis of :new with missing sparam

Similar to #46693, but for :new, rather than getfield.
Unfortunately, this is somewhat limited as we don't really
have the ability to encode type equality constraints in the
lattice. In particular, it would be nice to analyze:

```
struct Foo{T}
    x::T
    Foo(x::T) where {T} = new{T}(x)
end
```

and be able to prove this nothrow. If it's really
important, we could probably pattern match it, but
for the moment, this is not easy to do. Nevertheless,
we can do something about the similar, but simpler pattern

```
struct Foo{T}
    x
    Foo(x::T) where {T} = new{T}(x)
end
```

which is what this PR does.

Author: Keno

base/compiler/optimize.jl

@@ -254,16 +254,31 @@ function stmt_effect_flags(lattice::AbstractLattice, @nospecialize(stmt), @nospe
             nothrow = is_nothrow(effects)
             return (consistent, effect_free & nothrow, nothrow)
         elseif head === :new
-            typ = argextype(args[1], src)
+            atyp = argextype(args[1], src)
             # `Expr(:new)` of unknown type could raise arbitrary TypeError.
-            typ, isexact = instanceof_tfunc(typ)
-            isexact || return (false, false, false)
-            isconcretedispatch(typ) || return (false, false, false)
+            typ, isexact = instanceof_tfunc(atyp)
+            if !isexact
+                atyp = unwrap_unionall(widenconst(atyp))
+                if isType(atyp) && isTypeDataType(atyp.parameters[1])
+                    typ = atyp.parameters[1]
+                else
+                    return (false, false, false)
+                end
+                isabstracttype(typ) && return (false, false, false)
+            else
+                isconcretedispatch(typ) || return (false, false, false)
+            end
             typ = typ::DataType
             fieldcount(typ) >= length(args) - 1 || return (false, false, false)
             for fld_idx in 1:(length(args) - 1)
                 eT = argextype(args[fld_idx + 1], src)
                 fT = fieldtype(typ, fld_idx)
+                # Currently, we cannot represent any type equality constraints
+                # in the lattice, so if we see any type of type parameter,
+                # there is very little we can say about it
+                if !isexact && has_free_typevars(fT)
+                    return (false, false, false)
+                end
                 eT ⊑ₒ fT || return (false, false, false)
             end
             return (false, true, true)

test/compiler/effects.jl

@@ -655,3 +655,13 @@ end # @testset "effects analysis on array ops" begin
 
 # Test that builtin_effects handles vararg correctly
 @test !Core.Compiler.is_nothrow(Core.Compiler.builtin_effects(Core.Compiler.fallback_lattice, Core.isdefined, Any[String, Vararg{Any}], Bool))
+
+# Test that :new can be eliminated even if an sparam is unknown
+struct SparamUnused{T}
+    x
+    SparamUnused(x::T) where {T} = new{T}(x)
+end
+mksparamunused(x) = (SparamUnused(x); nothing)
+let src = code_typed1(mksparamunused, (Any,))
+    @test count(isnew, src.code) == 0
+end