Fix 3 related issues on move_base action recovering

[corot]

• move_base action get's stuck if recovery behavior patience is exceeded
• If a recovery behavior fails, calling the next immediatly breaks simple action client
• If the last recovery behavior fails, we return that as move_base error instead of the error that triggered recovering

I was debugging the first, but realized the 2 other on the process

The solution for the first is to treat patience exceeded the same way as behavior failure. Then I realized the other 2 errors.
The solution for the 2nd is rather hackish, but a proper solution would require to implement the move_base action, for which I don't have time. In any case the previous code was incorrect.

[siferati]

Suggested change

	// recovery failed but we still have behaviors; we cannot call the next here (simple action client
	// will complain about transition to ACTIVE from DONE), so we try replanning again (wrong, but works)
	case actionlib::SimpleClientGoalState::SUCCEEDED:
	// recovery failed but we still have behaviors; we cannot call the next here
	// (simple action client will complain about transition to ACTIVE from DONE)
	// so we try replanning again (wrong, but works)
	[[fallthrough]];
	case actionlib::SimpleClientGoalState::SUCCEEDED:

S: add [[fallthrough]] attribute to make it clear what the comment is referring to.
I was a bit confused at first until I noticed it was referring to the lack of a break statement 😅
I think this requires C++17 though, so if you don't want to change CMakeLists.txt then maybe add it commented out

S: better text line breaks