Update wasmer to 4.3

[chipshort]

closes #1628

Wasmer bumped their module version, so no need to bump our MODULE_SERIALIZATION_VERSION

Looks like the coverage job is running out of memory again.
@webmaster128 Can we bump that to a larger resource class?

The package_vm_windows job was also failing. I did a full cargo update to fix that (I couldn't easily figure out what package caused the linker problems. If you have tips, please tell me), followed by downgrading wat to 1.207.0 (instead of 1.208.1) to avoid their Rust 1.76 requirement.

[webmaster128]

@webmaster128 Can we bump that to a larger resource class?

Let's try to reduce the parallelization first. Cargo probably gets a value of 32 or 64 when asking the system for CPU count. For one other job we already have this:

    environment:
      # Limit the number of parallel jobs to avoid OOM crashes during doc testing
      RUST_TEST_THREADS: 8

[aumetra]

The package_vm_windows job was also failing. I did a full cargo update to fix that

Thought for a second that it might have been the cache, but can't be because the lockfile changed during the Wasmer upgrade.. weird.

[aumetra]

Wanna try to plug in an alternative linker since this seems to die on the ld invocation? mold claims to have a lower RSS than gold and lld

[aumetra]

Otherwise I'd even say let's go for -C instrument-coverage instead of gcov. Personally I had issues with OOM using tarpaulin and even gcov (IIRC) but never with cargo-llvm-cov

[chipshort]

Otherwise I'd even say let's go for -C instrument-coverage instead of gcov

But isn't that what is failing here? The compilation with -C instrument-coverage?
I think a different linker is worth a try.

[aumetra]

But isn't that what is failing here? The compilation with -C instrument-coverage?

NVM yeah, you're right. I was mixing up grcov and gcov lol. Would be nice if they came up with more distinct names next time

[chipshort]

Had to switch the coverage job to alpine 3.19 in order to get a somewhat recent mold version, but it seems to help with the memory usage

[webmaster128]

I don't have strong opinions which coverage tool to use. But if we change it, let's please do that in a separate PR

[aumetra]

Yep, definitely. But it seems like using a different linker did the trick already, so this looks good to go