Don't assume parameter presence in hierarchical likelihood

[bfarr]

The posterior predictive checks in hierarchical_likelihood() assume mass_1 and mass_ratio are present in pedata and injdata, e.g.,

GWInferno/gwinferno/pipeline/analysis.py

Lines 380 to 384 in 4460e22

	if posterior_predictive_check:
	if param_names is not None and injdata is not None and pedata is not None:
	pe_weights = jnp.exp(logpe_weights)
	inj_weights = jnp.exp(loginj_weights)
	cond = jnp.less(pedata["mass_1"], m1min) | jnp.greater(pedata["mass_1"], mmax)

which shouldn't be required.