[soukhova]

Second part of the fixes for issue #66 (subject: doubly-constrained accessibility).

Previously, the doubly-constrained iterative proportional fitting procedure (IPF) was run once across all transportation systems. This caused the land_use object to be integrated into the same run as many times as there were unique 'grouped' and 'multi-decay' parameters. This PR updates the logic such that each unique group or decay parameter is treated as an independent land-use-transportation-system, i.e., the function is run on each unique group and the the results are combined afterwards.

Changes:

• added run_standalone_doubly_ipf()
• added run_grouped_doubly_ipf()
• run doubly-constrained IPF separately for each group_by combination
• run doubly-constrained IPF separately for each decay_function_arg
• preserve grouping columns in output
• preserve decay_function_arg in output
• updated docs to reflect current doubly-constrained behaviour

Also added tests covering -- all passed on the 3x3 toy example, function continues to pass on existing tests

• grouped results equal standalone runs
• multiple decay values equal separate runs
• grouping columns are preserved
• grouped & multi-decay integration behaviour

As well, this implementation builds on the existing doubly-constrained fixes in dev mentioned in #67 (ensuring ID-based impedance matrix match and removal of the balancing-factor correction).

Closes #66 (and maybe #67 ?)