Fix TypeError for pGWP/pGTP with IPCC fallback functions

[TimoDiepers]

Problem

Calculating pGWP (or pGTP) on an inventory that contains a GHG outside the Watanabe module fails:

TypeError: create_generic_characterization_function.<locals>.characterize_generic()
got an unexpected keyword argument 'time_varying_re'

_characterize_pgwp passed time_varying_re to every flow's characterization function. With fallback_to_ipcc=True (the default), flows that Watanabe doesn't cover - CO and the GHGs built from decay_multipliers.json - are mapped to IPCC AR6 functions with signature (series, period, cumulative), which have no such parameter. In practice this hits any ecoinvent-based pGWP calculation, since the usual climate change methods characterize more than CO2/CH4/N2O.

Fix

_characterize_prospective_radiative_forcing already had this check inline. It is now the shared helper _apply_characterization_function, which passes time_varying_re only to the Watanabe functions and is used by the pGWP, pGTP and prospective radiative forcing paths alike.

Tests

New tests/test_characterize_prospective_fallback.py: a mixed characterization function dict (prospective CH4 + IPCC characterize_co + a generic decay-series function) across the metrics pGWP, pGTP and prospective_radiative_forcing, with time_varying_re both False and True. It reproduces the TypeError without the fix and asserts that all three flows end up characterized. Full suite: 100 passed.

Note on the file name: test_prospective.py replaces dynamic_characterization in sys.modules with a stub at import time, so any test module importing the real package has to sort before it alphabetically. That is noted in the new module's docstring, but it is a landmine worth removing separately.

Not addressed here

For IPCC fallback gases, pGWP divides an AR6 AGWP by a Watanabe AGWP for CO2, i.e. a mixed reference. Whether these gases should instead use the IPCC CO2 reference is a modelling decision, left out of this fix.

[TimoDiepers]

Pushed a second, related usability fix onto this branch: the RuntimeError raised when a prospective metric is calculated without a scenario.

Old message:

No scenario set. Call prospective.set_scenario(iam, ssp, rcp) first.

That assumes the reader knows what prospective is - but users reaching this through bw_timex have never imported dynamic_characterization themselves, and it doesn't say which arguments are valid. New message:

No scenario set.

The prospective characterization factors (Watanabe et al. 2026) used for the metrics
"pGWP", "pGTP" and "prospective_radiative_forcing" depend on a future scenario, so you
have to choose one before calculating them - also if you calculate them through another
package, such as bw_timex:

    from dynamic_characterization.prospective import set_scenario
    set_scenario(iam="IMAGE", ssp="SSP1", rcp="2.6")

This is done once per Python session and applies to all following calculations.

Each IAM comes with one SSP: IMAGE-SSP1, MESSAGE-SSP2, AIM-SSP3, GCAM4-SSP4,
REMIND-SSP5. The available RCPs are "2.6", "4.5", "6.0" and "8.5", but not for every
IAM - see dynamic_characterization.prospective.VALID_SCENARIOS for the full list.

Covered by test_get_scenario_without_set_message_is_actionable; the existing match="No scenario set" tests still pass. Happy to split this into its own PR if you'd rather keep #29 to the TypeError.