fix(sem): avoid no-return false positives

[saem]

Summary

no-return analysis now no longer confuses void branches for true
no-returns.

Details

No-return analysis now dives into the trees of if/try/block/case
statements in order to determine if they are in fact no-return
scenarios. This is only done for the statement variety.

The updated analysis besides being more correct, in that it avoids false
positive no-return identification, it still allows for a number of false
negatives. Chiefly that unstructured exits ( return/break/raise ) are
only considered in the trailing position of statement lists and not part
way through. This is considered acceptable as there is an overall
improvement in accuracy and it should cover most code.

This includes a fix for an NPE crash in discard-or-use error reporting.

[zerbina]

Looks good to me. Using recursion might make the analysis a tad easier to understand/read, but endsInNoReturn will be obsoleted by a proper void type anyway.

[zerbina]

/merge

[github-actions]

Merge requested by: @zerbina

Contents after the first section break of the PR description has been removed and preserved below:

---

Notes for Reviewers

• Found via Loony CI and reduced thanks to @alaviss

[saem]

Thanks for the review.

Looks good to me. Using recursion might make the analysis a tad easier to understand/read, but endsInNoReturn will be obsoleted by a proper void type anyway.

I did think about it, but already felt it's a bit pricey (cost to benefit) as is and didn't want to push it any further than I already have.