feat: improve rust codegen for wasm target

[bluenote10]

This PR improves the Rust codegen for targeting wasm, following up on ideas from #1118.

Basically now the generated code works for both non-wasm and wasm compilation. (My requirement was to have the same generated code work under both.)

To demonstrate, the generated code now looks like this:

#[cfg(not(target_arch = "wasm32"))] // Compile ffi bindings only on non-wasm targets
mod ffi {
    use std::os::raw::c_float;
    // Conditionally compile the link attribute only on non-Windows platforms
    #[cfg_attr(not(target_os = "windows"), link(name = "m"))]
    unsafe extern "C" {
        pub fn remainderf(from: c_float, to: c_float) -> c_float;
        pub fn rintf(val: c_float) -> c_float;
    }
}
fn remainder_f32(from: f32, to: f32) -> f32 {
    #[cfg(not(target_arch = "wasm32"))] // non-wasm targets use ffi bindings
    unsafe { ffi::remainderf(from, to) }
    #[cfg(target_arch = "wasm32")] // wasm relies on libm
    libm::remainderf(from, to)
}
fn rint_f32(val: f32) -> f32 {
    #[cfg(not(target_arch = "wasm32"))] // non-wasm targets use ffi bindings
    unsafe { ffi::rintf(val) }
    #[cfg(target_arch = "wasm32")] // wasm relies on libm
    libm::rintf(val)
}

This follows @mitchmindtree's idea in #1118 (comment) and works for both use cases:

• native targets: They use the ffi and linking to lm as before.
• wasm target: They delegate to libm, a pure Rust crate offering the same functionality. Of course, wasm users need to add libm as a dependency.

---

Note that this makes the -rnlm/--rust-no-libm command line args (and gRustNoLibm internally) more or less obsolete. Removing it would be a breaking change though for someone expecting these command line arguments to exist. I didn't remove it yet, but if the breaking change is considered minor enough, we could do so in a follow-up. This would be more or less the inverse of ad2e156.

[sletz]

@crop2000 You added -rnlm/--rust-no-libm right ? So can we remove them ?

[bluenote10]

@crop2000 You added -rnlm/--rust-no-libm right ?

@sletz If I remember correctly @crop2000 had added the -rnt/--rust-no-faustdsp-trait (which is independent), but I believe -rnlm/--rust-no-libm was added by you in ad2e156 more or less as a quick fix for #1118.

[sletz]

OK, so better wait for @mitchmindtree feedback then ?

[bluenote10]

OK, so better wait for @mitchmindtree feedback then ?

My idea was to merge this approach and release it so that it's easy for people to use it and get rid of their --rust-no-libm usages first. Then we can still make the decision whether to remove it at a later point based on feedback.

[sletz]

OK done, thanks.