Resolve "Speed up LabelledDataGroup "all" selection"

This MR improves the speed of the LabelledDataGroup selection. In particular the shortcuts for special cases (select single member, select all) are provided/improved. Moreover the data preparation is made more efficient for the merge combination method. Along with that the selection methods are restructured and separated to make everything more transparent.

A new combination method, auto, is added, which automatically chooses between merge and try_concat based on the datatype.

Details

• Implement the _select_all_merge shortcut:
  • Directly iterate over the group members and try to merge them instead of (building and) using the member map
• Add the auto combination method (and make it the default option):
  • merge combination changes the dtype to float when adding nans due to missing values, whereas concat does not. Therefore, choose try_concat if dtype is int and merge otherwise.
• Speed up selection when combining via merge:
  • Use .where on the pre-selected member map to get all coordinates for the respective containers (avoiding to select the same group member multiple times).
• Extend and restructure tests to cover as many selection scenarios as possible

Anything to double-check?

• Selection interface transparent and consistent?

Can this MR be accepted?

• Implementation ready
• Tests added or adjusted
• Documentation extended or updated (Adapted the docstrings)

• Code quality
  • Changes follow coding guidelines
  • Checked code coverage on new and adjusted code

• Ready for merging
  • Pipeline passes without warnings
  • History cleaned-up or squash option set
  • Changelog entry added
  • Version number bumped to v0.17.0a0
  • Reviewed & approved by @

Related issues

Closes #93 (closed)