Update data representation management at stage and pipeline level and fix service docs

[thehrh]

Key changes and fixes:

• remove consistency checks on apply_modes and prepare defining sum as default translation mode for certain container keys -> address in another PR (see discussion Re-define and document supported combinations of calc_mode and apply_mode in pipeline configurations #933)
• expose whether service overrides setup_function, compute_function, or apply_function via has_setup, has_compute, and has_apply function -> default deduction of allowed representations at Stage level
• services with restrictions on their data representations now explicitly declare which ones they support
  • removed need for given service to specify each dict value as a list (e.g. None instead of [None])
• removed instances of superfluous representation setting, when the Pipeline already takes care of that
• a recently introduced dangerous bug is fixed that allowed running with a representation set to None even though the corresponding function is implemented (see new tests in core.pipeline module)
• representation consistency is now validated separately in setup(), compute(), and apply() functions in case the Stage somehow became misconfigured (unlikely to have any significant performance impact on realistic pipelines IMO)
• give pisa_tests/test_services.py the ability to generate a service implementation reference (table/matrix) in markdown & csv format, and include said markdown table in notebook pisa_examples/pisa_modes.ipynb

Documentation improvements:

• fixed and standardised docstring formatting across all services (in particular parameter & notes—expected container keys—sections)
• clean up "PISA Pi" mentions
• updated copyright years