Allow empty slice when loading log with sublogs

[jl-wynen]

Loading an NXlog that contians sublogs with x[:] raises an exception because we don't support slicing sublogs. But in this case, it should behave the same as x[()] or x[...] for 1D data. So this PR handles it this way.

I ran into this in ESSreduce when passing the default TimeInterval (slice(None, None, None)) to a loader for an NXlog.

[SimonHeybrock]

Why is the branch below that does elif isinstance(select, int | sc.Variable) or isinstance(select, slice): not sufficient?

[jl-wynen]

Because that returns a 1d slice which leads to this error downstream:

     59 def read_children(self, sel: ScippIndex) -> sc.DataGroup:
     60     # Sublogs have distinct time axes (with a different length). Must disable
     61     # positional indexing.
     62     if self._sublogs and ('time' in to_canonical_select(list(self.sizes), sel)):
---> 63         raise sc.DimensionError(
     64             "Cannot positionally select time since there are multiple "
     65             "time fields. Label-based selection is not supported yet."
     66         )
     67     dg = super().read_children(sel)
     68     for name, field in self._sublog_children.items():

DimensionError: Cannot positionally select time since there are multiple time fields. Label-based selection is not supported yet.

scippnexus/src/scippnexus/nxlog.py

Lines 59 to 70 in 1ca907d

	def read_children(self, sel: ScippIndex) -> sc.DataGroup:
	# Sublogs have distinct time axes (with a different length). Must disable
	# positional indexing.
	if self._sublogs and ('time' in to_canonical_select(list(self.sizes), sel)):
	raise sc.DimensionError(
	"Cannot positionally select time since there are multiple "
	"time fields. Label-based selection is not supported yet."
	)
	dg = super().read_children(sel)
	for name, field in self._sublog_children.items():
	dg[name] = field[sel]
	return dg

Alternatively, we could change the check here.

[SimonHeybrock]

Right, so this specifically fixes a bug when the ESS-style "sublogs" are present (which are not in the standard, iirc)? A more targeted fix might indeed make sense, but I don't mind too much.

Can you update the PR title/description?

I think this also means that the new tests do not really do much that didn't work before, as they contain no sublogs as far as I can see. The removed line in test_log_with_connection_status_raises_with_positional_and_label_indexing is I think the test you actually need, except that it would now not raise.

[jl-wynen]

Updated