Shared: Do not use @kind graph for CFG test output

[hvitved]

The original motivation for using @kind graph was to be able to visualize CFGs in VS Code. However, now that we have View CFG, this is no longer needed. Moreover, @kind graph often has spurious reorderings of the output, and there are some known limitations (such as missing multi-edges, IIRC), so it is not well suited for qltest.

[michaelnebel]

C# LGTM!

[redsun82]

Seems reasonable, thanks for jumping on that!

The only comment I have is that the expected files remain somewhat hard to check for a human (more or less as before). But small diffs (as will usually be the case in a PR) are probably manageable.

[hvitved]

The only comment I have is that the expected files remain somewhat hard to check for a human (more or less as before). But small diffs (as will usually be the case in a PR) are probably manageable.

Diffs are, IMO, easier to check in the non-@kind graph format.

[redsun82]

The only comment I have is that the expected files remain somewhat hard to check for a human (more or less as before). But small diffs (as will usually be the case in a PR) are probably manageable.

Diffs are, IMO, easier to check in the non-@kind graph format.

I agree that typical PRs are probably easier to review now. The full expected file though remains pretty much imperscrutable 😅