Handle pseudo return nodes in Node.of_id and improve their location

[sim642]

Closes #997.

Changes

1. Adds Cilfacade.pseudo_return_stmt_sids, which maps pseudo return node statement IDs to the statements themselves. This allows Cilfacade.find_stmt_sid, which is used by Node.of_id to work for pseudo return nodes.
2. Previously pseudo return nodes were given the location of the entire function definition, which starts already before the function's return type. This PR changes their location to be the end location of the function definition (with 0-width range), corresponding to the }. Since Locator uses nodes from the CFG, this already includes pseudo return nodes, but now with more reasonable locations for both lookup and display.

There is one unexpected point though as indicated by the change to the cram test: our LLoC counting can increase because the pseudo return node at its new location is on a separate line. Not sure if this weird or fine.

[michael-schwarz]

Btw everyone, this is no #1000 🎉

[sim642]

There is one unexpected point though as indicated by the change to the cram test: our LLoC counting can increase because the pseudo return node at its new location is on a separate line. Not sure if this weird or fine.

It might be possible to avoid this change by making dead code statistics ignore pseudo return nodes. Maybe that's what we should be doing because they don't correspond to anything real in the user code anyway?

[stilscher]

I think that is a good idea.