Regression: No augmented forward pass found for _RNvCsbpInmopFggX_7___rustc19___rust_alloc_zeroed(__rustc::__rust_alloc_zeroed)

Here's another regression, autodiff on Karpathy's gpt2 doesn't work anymore. The mwe is manually minimized/translated from Rust
[mwe.ll.txt](https://github.com/user-attachments/files/22267473/mwe.ll.txt)
This is the semi-automatically minimized version:

```
target datalayout = "e-m:e-p270:32:32-p271:32:32-p272:64:64-i64:64-i128:128-f80:128-n8:16:32:64-S128"
target triple = "x86_64-unknown-linux-gnu"

define fastcc {} @_ZN5llmrs6main_017ha3c193b20e45989fE() {
%1 = call {} (...) @__enzyme_autodiff_ZN5llmrs16gpt2_ad_backward17hdd0c6bf2325259a6E(ptr @gpt2_forward, metadata !"enzyme_dup", ptr null, ptr null, metadata !"enzyme_const", ptr null, metadata !"enzyme_const", ptr null, metadata !"enzyme_const", ptr null, metadata !"enzyme_const", i64 0, metadata !"enzyme_const", i64 0, metadata !"enzyme_dup", ptr null, ptr null)
ret {} %1
}

define void @gpt2_forward(ptr %0, ptr %1, ptr %2, ptr %3, i64 %4, i64 %5, ptr %6) {
call void @matmul_forward(ptr %0)
ret void
}

define void @matmul_forward(ptr %0) {
call void @_RNvCsbpInmopFggX_7___rustc35___rust_no_alloc_shim_is_unstable_v2()
%2 = tail call ptr @_RNvCsbpInmopFggX_7___rustc19___rust_alloc_zeroed(i64 noundef range(i64 1, 0) 32, i64 noundef range(i64 1, -9223372036854775807) 4)
%3 = load float, ptr %2, align 4
store float %3, ptr %0, align 4
ret void
}

; Function Attrs: nounwind nonlazybind
declare void @_RNvCsbpInmopFggX_7___rustc35___rust_no_alloc_shim_is_unstable_v2() unnamed_addr #1

; Function Attrs: nounwind nonlazybind allockind("alloc,zeroed,aligned") allocsize(0)
declare ptr @_RNvCsbpInmopFggX_7___rustc19___rust_alloc_zeroed(i64, i64)

declare {} @__enzyme_autodiff_ZN5llmrs16gpt2_ad_backward17hdd0c6bf2325259a6E(...)

```

Also:
```
/home/manuel/prog/llvm21/build/bin/llvm-cxxfilt _RNvCsbpInmopFggX_7___rustc19___rust_alloc_zeroed
__rustc::__rust_alloc_zeroed
```
And
```
enzyme/Enzyme/LibraryFuncs.h
58: if (name == "__rust_alloc" || name == "__rust_alloc_zeroed")
210: if (funcName == "calloc" || funcName == "__rust_alloc_zeroed")

enzyme/Enzyme/GradientUtils.cpp
9476: if (allocationfn == "__rust_alloc" || allocationfn == "__rust_alloc_zeroed") {

enzyme/test/Enzyme/ReverseMode/initializeretrust.ll
11: %call = tail call i8* @__rust_alloc_zeroed(i64 %mul, i64 4)
20:declare dso_local noalias i8* @__rust_alloc_zeroed(i64, i64)
85:; CHECK-NEXT: %"call'mi" = tail call noalias nonnull i8* @__rust_alloc_zeroed(i64 %mul, i64 4)
86:; CHECK-NEXT: %call = tail call i8* @__rust_alloc_zeroed(i64 %mul, i64 4)
```

Not sure why the Enzyme logic regressed?