reference implementation fails to reject pulses with a proportion of zero

[grahamgower]

We have no test cases with a pulse proportion of zero, and the reference implementation erroneously accepts such pulse proportions. Reported by @molpopgen over at popsim-consortium/demes-python#576

[grahamgower]

It looks like this issue also affects ancestry proportions. I.e. the following is accepted but should not be.

time_units: generations
demes:
- name: a
  epochs:
  - {start_size: 100, end_time: 50}
- name: b
  epochs:
  - {start_size: 100, end_time: 50}
- name: c
  ancestors: [a, b]
  proportions: [0.0, 1.0]
  start_time: 100
  epochs:
  - {start_size: 100}

[grahamgower]

Fixed by #167.