[NativeAOT-LLVM] map runtimeimports to their native name for ldftn

[yowl]

This PR partially addresses a problem with ldftn for RuntimeImports, i.e. from

runtimelab/src/coreclr/nativeaot/Runtime.Base/src/System/Runtime/RuntimeExports.cs

Line 455 in 3dd23bf

internal static unsafe IntPtr RhGetRuntimeHelperForType(MethodTable* pEEType, RuntimeHelperKind kind)

Right now these are reported as undefined as they are not translated to the native names:

warning: undefined symbol: S_P_CoreLib_System_Runtime_InternalCalls__RhpNewArray (referenced by top-level compiled C/C++ code)

Emscripten will create imports for these in the wasm module, and stub them in JS. However a WASI host like wasmtime cannot satisfy these imports and just refuses to start. This PR at least lets the module run in a WASI host, although if these function pointers were ever called, they might be treated as managed calls, get a shadow stack arg, and fail. I propose to not solve that in the IL->LLVM compilation, but leave it until it's hit in the RyuJIT compilation.

There is still one undefined symbol preventing wasmtime starting, dotnet_browser_entropy, but that can be tackled in it's own PR.

cc @SingleAccretion

[yowl]

Oh we are hitting the hash problem emscripten-core/emscripten#16164 We'll have to wait for that to be resolved.

[yowl]

The emscripten bug is fixed so this can be reviewed now. Thanks.

[jkotas]

propose to not solve that in the IL->LLVM compilation, but leave it until it's hit in the RyuJIT compilation.

Ok, sounds reasonable.

[jkotas]

propose to not solve that in the IL->LLVM compilation, but leave it until it's hit in the RyuJIT compilation.

Ok, sounds reasonable.