Fix an issue with unsorted deleteat

[lkdvos]

There was a small issue when the trace indices were generated in unsorted order, which made deleteat complain. This should resolve that.

[Jutho]

I missed this PR. Do you happen to remember an example that triggered this? Might be good to add to the tests.

[lkdvos]

I think all you need is more than one pair of trace indices that are not sorted. It's only used in the cutensor implementation, so it might not have run on CI because of that, although I'm not so sure that we have something that triggers it.

[Jutho]

I am actually confused why deleat complains. It sorts the indices internally exactly in order not to run into such problems:
https://github.com/Jutho/TupleTools.jl/blob/2a6b88c8153a0a0d83ec3cde60b9937d9122402b/src/TupleTools.jl#L119

[Jutho]

Should this be part of the 5.0.2 release? I am not opposed to it, but if this fix is necessary, it points to a bug in TupleTools that I would like to understand.

[lkdvos]

I'm fine with having this separate if necessary, could be 5.0.3, I should have access to my laptop again starting next week so I can check then to make sure I wasn't just doing something silly.

(I should apparently read my own advice more carefully: including a MWE, a Stacktrace etc, would have resolved this issue already...)

[lkdvos]

I double-checked, actually it is here that the issue stems from: Ainds is a vector, rather than a tuple, so this does not go via TupleTools. I will try the following:

• make sure trace_indices returns sorted indices
• add a test for this

Edit: I cannot fix this in trace_indices, and was too quick in suggesting that. I do think the current solution is the easiest, but will still try and add a test.