Get rid of in-substrate usages of core_intrinsics feature

[pepyakin]

Removes all instances of usage core_intrinsics. In two cases they were really useless. In the third case, we could substitute it with a call to core::arch::wasm32::unreachable which is stable.

One little step towards #1252

The behavior hasn't changed, thus no bumping of runtime is required.

[pepyakin]

The CI failure implies that the nightly used is rather old.

[bkchr]

But why does it work for the main build where we build the wasm blobs? Hmm.

[pepyakin]

My hypothesis is that the gitlab-test-full-crypto-feature job uses an outdated nightly. I am not sure though, @gabreal might be able to confirm this.

[pepyakin]

Ok, failure is in fact legit: core::arch::wasm32 is only available for wasm32 and for some reason we compile sr-io for no_std but not wasm32 (probably, SGX?).

I think I am inclined to disable panic_handler and alloc_error_handler since they are disabled for those use-cases anyway.