Fix deserialization of MatGroup

[lgoettgens]

Resolves #5882 by adding an upgrade script.

I am not quite sure if I assembled the upgrade script in the right way, as there was no pure renaming since we have upgrade_recursive. In particular, I am not sure if upgrade_recursive needs to be called.

As the corresponding renaming was already released with Oscar 1.7.0, IMO we should backport this asap. (cc @benlorenz @aaruni96 )

[lgoettgens]

Ok, so it seems that rename_types does not work anymore for recent-ish upgrades... @antonydellavecchia could you take this from here?

[benlorenz]

I added three examples here: oscar-system/serialization-upgrade-tests#13
One of these is from the leech pairs database, the other two are saved with Oscar 1.6.0.

I set DO NOT MERGE to make sure this is not merged before the test files.

[ThomasBreuer]

Concerning the commit "dont recurse into :data section for type rename":
I had copied the code from the update script 1.2.0.jl, the "youngest" such script that involves rename_types.
So either the recursion was already unnecessary in that script, or the data format has been changed since then.

[benlorenz]

Concerning the commit "dont recurse into :data section for type rename": I had copied the code from the update script 1.2.0.jl, the "youngest" such script that involves rename_types. So either the recursion was already unnecessary in that script, or the data format has been changed since then.

In theory, types names should not be within the data section, but we had some cases where this was violated. There were some fixes regarding this at some point but I don't remember when exactly.
So it is possible that this was not necessary but it is also possible that this was fixed after 1.2.
But I think it is more likely that this block was never necessary, removing it (from the 1.2 upgrade script) doesn't seem to break anything. But still, to be safe I would just keep it as it is.

[lgoettgens]

Thanks @benlorenz! I would approve, but that's not possible because this is formally my own PR. Once the hash has been updated to the merge commit of oscar-system/serialization-upgrade-tests#13, this is good to go from my POV.