Releaxed trunc corner cases

[ppenzin]

Chatted with @rossberg offline, and we landed on:

• For relaxed cases where the hardware differences exceed 2-4 well defined cases, the core spec and the tests should under specify/test to be maximally inclusive of future implementations (ex: potential new hardware backend with slightly different semantics)
• Web based implementations can and should still test edge cases but that can live elsewhere which we could decide - WPT tests are an option
• one thing that would be good to discuss though is how to document the nuances of the special cases. These could be in notes for the instruction, an appendix, or some user facing documentation like the intrinsics documentation are also a possibility. If the next edition of the subgroup meeting has slots available, I'd like to add this to the agenda so we can resolve it.

Originally posted by @dtig in https://github.com/WebAssembly/relaxed-simd/issues/144#issuecomment-2719169220

[ppenzin]

Next instance is tomorrow, but we have agenda slots available.

[ppenzin]

An example of what is missing from current relaxed trunc tests: https://github.com/WebAssembly/relaxed-simd/pull/140 Though if the path is to try to generalize even further then that might not be enough.

[yurydelendik]

There is also https://github.com/WebAssembly/relaxed-simd/blob/main/proposals/relaxed-simd/Entropy.md . I'm not sure that status this document has, but "Entropy exposed" fields must the re-evaluated for trunc instructions.