Allocate empty attribute arrays for detectors

[angus-g]

When we allocate/unpack detectors from Zoltan, some code expects that their attribute arrays (used for particles) are at least allocated, even if length 0. Ensuring this is done avoids potential segfaults when these arrays are unconditionally accessed.

Thanks @stephankramer for the detailed analysis of this issue! I think this is probably the right place for this allocation? Of course, the other approach would be to figure out where attribute_size is being passed into unpack_detector as a non-optional on a detector. (As an aside, that subroutine looks like it could just take a non-optional attribute_size and then the logic wouldn't be exactly duplicated in both branches...)

Closes #317.

[angus-g]

Hmm, not sure why tests are failing in the Docker configuration here. Is it because I opened the PR against a branch on my own fork?

[Patol75]

Answer here I believe: https://github.com/FluidityProject/fluidity/pull/303#issuecomment-836654951

[tmbgreaves]

@angus-g - yes, sorry about that - I've not yet figured out a clean way to fix PRs from outside the main repo. The problem (I think) is that it's indeterminate where to pull docker secrets from, or possibly a security issue for remote forks to be able to use secrets from the Fluidity repo, so the docker login fails. We tried replicating the secrets in the PR source repo but that didn't work either.

Might you be able to push your branch to the main Fluidity repo and PR from that please?

[tmbgreaves]

Thank you very much for the fix here! Additionally, I think (from @drhodrid 's comment) that @Cmath2 is no longer working on Fluidity; might you be able to request a different reviewer when reopening the PR please?