MD return values -- what should they be

[zerothi]

The problem is that we need to converge the returned values in some way.

Currently some methods returns

• lists
• collection
• geometrycollection

Now what should we do about multiple returns. See below for two ideas, which would we prefer.

Things to consider:

• easy to work with

• get performant code (ideally they could be (md, ...) arrays)

• consistency across the outputs

          I've addressed all comments except this one as I am not sure we want/need to unpack. For instance, the current behaviour seems logical to me:
  

traj = stdoutSileVASP(f).read_trajectory[:3]()
for i in range(3):
    print(traj[i].force)
# or
for step in traj:
    print(step.force)

You would prefer this?

traj = stdoutSileVASP(f).read_trajectory[:3]()
for i in range(3):
    print(traj.force[i])
# or
for F in traj.force:
    print(F)

Originally posted by @tfrederiksen in https://github.com/zerothi/sisl/pull/573#discussion_r1235576392

List of actions:

• [x] #589 remove Collection and friends, users now need to manually write many geometries
• [ ] Dataset attempt