add mass distributions from arXiv 2111.03498

This patch adds an additional Mass1Distribution and a few ComponentMassPairing object. These are inspired by arXiv:2111.03498 (Bridging the Gap).

• ButterworthNotchNPiecePowerLawMass1 (tested via test/test-mass1-distributions)
  • adds low-pass, high-pass, and notch filters on top of the NPiecePowerLawMass1 infrastructure. The naming convention for the filter parameters is slightly different, but the functional forms are Butterworth filters as described in arxiv:2111.03498
• BinnedPowerLawAsymmetricMassRatioPairing`` (tested via test/test-mass-pairing-distributions`)
  • a pairing function based on the mass ratio that "switches" the exponent depending on the secondary mass. This switching is discontinuous (pairing function is not continuous as a function of mass2), and this leads to discontinuous marginal distributions. This pairing function is what was used in arxiv:2111.03498.
• ContinuousBinnedPowerLawAsymmetricMassRatioPairing (tested via test/test-mass-pairing-distributions)
  • an extension of BinnedPowerLawAsymmetricMassRatioPairing that includes an additional factor to the "high-secondary mass" option that renders the entire pairing function continuous in mass2. This also makes the marginal distributions continuous. Note that this pairing function was not used in arxiv:2111.03498.

The output of the test scripts are included as comments to document the behavior of each object. Objects were tested for compatibility with both numpy and jax backends. Although the examples reported below come from only the numpy tests, both backends were shown to behave correctly.

This patch also picked up a few small changes to make SamplingDistribution._check_variates play nicely with more data formats.

/cc @amanda.farah