Proportional initializer fails when ratios is a C-length array and N == C

[JavadocMD]

The documented intention for the Proportional initializer is that you can provide either an NxC-shaped array OR a C-shaped array for the ratios argument. In the case of a C-shaped input, this should be broadcast to NxC (where the input is duplicated across the rows, the N-axis). This mostly works as expected except for the special case where N is equal to C (there are the same number of nodes as IPM compartments). The broadcasting happens in the wrong direction (across the C-axis), which eventually raises an error or returns unexpected results depending on the values used.

We should make sure to explicitly interpret 1-dimensional inputs as C-values when broadcasting.

While we're at it, the error messaging around invalid ratios forms could be improved.

Example 1: raises an error

from epymorph.kit import *

init.Proportional(ratios=[0.9, 0.1, 0.0]).with_context(
    scope=CustomScope(["A", "B", "C"]),
    ipm=ipm.SIRS(),
    params={
        "population": [100, 200, 300],
    },
).evaluate()

---------------------------------------------------------------------------
InitError                                 Traceback (most recent call last)
Cell In[3], line 9
      1 from epymorph.kit import *
      3 init.Proportional([0.9, 0.1, 0.0]).with_context(
      4     scope=CustomScope(["A", "B", "C"]),
      5     ipm=ipm.SIRS(),
      6     params={
      7         "population": [100, 200, 300],
      8     },
----> 9 ).evaluate()

File ~/Workspaces/Epymorph/epymorph/simulation.py:528, in SimulationFunctionClass.__new__.<locals>.evaluate(self, *args, **kwargs)
    526 @functools.wraps(orig_evaluate)
    527 def evaluate(self, *args, **kwargs):
--> 528     result = orig_evaluate(self, *args, **kwargs)
    529     self.validate(result)
    530     return result

File ~/Workspaces/Epymorph/epymorph/initializer.py:211, in Proportional.evaluate(self)
    209 if np.any(row_sums <= 0):
    210     err = "One or more rows sum to zero or less."
--> 211     raise InitError(err)
    213 pop = self.data(_POPULATION_ATTR)
    214 result = pop[:, np.newaxis] * (ratios / row_sums[:, np.newaxis])

InitError: One or more rows sum to zero or less.

Example 2: no error but bad results

from epymorph.kit import *

init.Proportional(ratios=[0.5, 0.3, 0.2]).with_context(
    scope=CustomScope(["A", "B", "C"]),
    ipm=ipm.SIRS(),
    params={
        "population": [100, 200, 300],
    },
).evaluate()

array([[ 33,  33,  33],
       [ 67,  67,  67],
       [100, 100, 100]])