Clear some forgotten fields in Triangulation::clear().

[bangerth]

In 1d, triangulations store maps from vertex index to manifold and boundary ids. We forgot to clear these fields in Triangulation::clear() (or rather the internal function clear() calls).

I don't think this has an effect that is observable because once you create another triangulation in the object, we add or overwrite the elements of this map, and when we access it we only ever access the elements we have just added or overwritten -- elements that just happen to still be there are simply never accessed, but they take up memory and we should really reset these maps in clear() as well.

I found this in #15689 where I actually am accessing all elements of these maps in order to translate vertex numbers, and I'm running into cases where some of the elements are invalid (because left-over from a triangulation we no longer store). This patch fixes this.

[tjhei]

Can you use .reset() instead? Afterwards, the comment will be redundant too.

[bangerth]

.reset() resets the pointer to a nullptr. What I need is to let the pointer point to an empty object.

Or are you suggesting to call .reset() on the pointed-to object? Since the object pointed to is a map, you would then mean to call .clear() on it?

[tjhei]

Sorry, you are right. Are they guaranteed to be non null? If yes, .clear () seems like a sensible thing to do or do you disagree?

[bangerth]

Yes, they are guaranteed to be non-null -- but one can also assert this :-)

I've pushed another patch that makes the modification you suggest (and that adds more assertions as well).