Improve data selection for diagnostics requiring a parent dataset

[bouweandela]

At the moment, many ECS diagnostic runs are failing because insufficient input data is available. This particular diagnostic needs 150 years of data from the start of the abrupt-4xCO2 experiment. We could reduce the number of failures by e.g. adding the option to specify a duration in RequireTimerange constraint or by adding a new constraint to specify this.

[bouweandela]

This may also be helpful for the TCR diagnostic.

For both these diagnostics, the number of failures could also be reduced by moving the temporal selection happening, e.g. in

climate-ref/packages/climate-ref-esmvaltool/src/climate_ref_esmvaltool/diagnostics/tcr.py

Lines 182 to 188 in 9f5541a

	recipe["datasets"] = get_child_and_parent_dataset(
	df[df.variable_id == "tas"],
	parent_experiment="piControl",
	child_duration_in_years=140,
	parent_offset_in_years=0,
	parent_duration_in_years=140,
	)

to the AddParentDataset constraint.

[bouweandela]

The function linked above should also be rewritten to take the child_experiment as an argument instead of the parent_experiment, because the parent experiment may differ per dataset, while the child_experiment is known because it is specified in the data requirements.