build: add -Wimplicit-fallthrough

[idryzhov]

There are currently no cases of implicit fallthrough in FRR. Adding this to be safe in the future.

[donaldsharp]

once ci finishes I'll push this in

[idryzhov]

Wow that's interesting. It looks like we actually have some implicit fallthrough but only OpenBSD build catches them. I tested locally on Ubuntu 20.04 and didn't see the errors. CI doesn't have errors on Ubuntu as well...
I'll look into it tomorrow.

[donaldsharp]

clang-15 reproduces this problem for me on my build machine

[idryzhov]

So yes, OpenBSD uses clang as a compiler and there were two problems with the current code:

1. /* fallthrough */-like comments are not portable as they are not supported by clang. I replaced all the comments with portable fallthrough; pseudo-keyword defined in compiler.h. It was already defined as _FALLTHROUGH; but I renamed it because our doc recommends fallthrough; and Kernel uses it as well.
2. clang doesn't allow the following:

   case X:
       do_something();
   default:
       break;
   

   I added explicit break; statements to affected places.

[ton31337]

LGTM

[donaldsharp]

I disagree that this should be a fallthrough to the below message. The message seems wrong to me.

[donaldsharp]

again the actions taken in the default seem to be off to me.

[donaldsharp]

Mark, Igor and myself spoke in slack and had this conversation:

a) This changes exposes some problems in the code from my perspective, but it is not Igor's fault for this exposure and as such should not be punished
b) The choice of a fallthrough was fundamentally wrong in more than a few places in the current code base. Some of these are extremely obvious how to fix and some are not.
c) How can we capture these and come back and fix them in the future?

Igor has agreed to go through and fix the easy ones. The harder ones I am tracking down the original implementer and asking them to fix them problems.

The easy one's should not be too much of a problem to go through and fix, all three of us have seen changes that need to be made and Igor should be positioning these changes soon.

Let's hold off pushing this in until Igor updates the PR again.

[idryzhov]

I cleaned up all places where the use of fallthrough; was not necessary/made the code more complex. The remaining question is whether fallthrough usage is correct in LDPD and NHRPD. Waiting the original implementer to confirm.