Fix bc_label loop in scalar_scheme_t

[timofeymukha]

There is an issue with how we parse bc_labels, which seems to so far simply gone undetected. By default, the size of the bc_labels array is NEKO_MSH_MAX_ZLBLS. However, when we pass bc_labels to the json_get, the array gets reallocated to the size as found in the case file. So, all the loops that parse bc_labels and iterate up to NEKO_MSH_MAX_ZLBLS are going through garbage. Now, most of these loops search for an exact string comparison, which will essentially never happen.

But for the scalar, we search for the first element of the label to be "d" (or "n", once Neumann is merged), and that can actually randomly be there for a garbage bc_label item, which will of course crash everything. Therefore, I fix the bc_labels to be an assumed shape array and iterate up to its size.

This will fix things in practice, but in principle we should do the same everywhere we loop through the bc_labels, and throw error if json_get reallocates to a size, which is larger than the NEKO_MSH_MAX_ZLBLS.

Question: why do we have limit on the bc_labels size at all?

[timofeymukha]

The reason for the max size came from the old name list format

OK, I see. So we can retire that now?

[njansson]

The reason for the max size came from the old name list format

OK, I see. So we can retire that now?

Not sure, maybe it's still used to define the zone array for a mesh

[timofeymukha]

Alright. I will open an issue so that we remember about this.