Update simple test

[loganek]

• don't assume the return value of the __wasi_thread_spawn API should be zero; assert on thread ID
• implement wasi_thread_start in ASM to avoid potential stack corruption

[yamt]

do you want to keep this while #19 has a similar coverage?

[loganek]

do you want to keep this while #19 has a similar coverage?

Good point; I think we can keep it and use it as an example for C.

[yamt]

do you want to keep this while #19 has a similar coverage?

Good point; I think we can keep it and use it as an example for C.

but this is now involving asm.

[loganek]

Yes, partially. However, most of that is C, so I still see the value of having that one test. We could possibly reduce the amount of asm code by having a small trampoline function written in C with asm inlines for setting stack pointer and force compile it with optimizations (IIRC, that is safe and won't generate code that uses stack pointer). But that will only limit the amount of asm code, won't eliminate it completely.