Missing idata attributes from jax/numba samplers

[ricardoV94]

Description

import pymc as pm

with pm.Model() as m:
    x = pm.Normal("x")
    idata1 = pm.sample(nuts_sampler="pymc")
    idata2 = pm.sample(nuts_sampler="numpyro")

print(idata1.posterior.attrs)
# {'created_at': '2024-04-17T08:06:27.074686',
#  'arviz_version': '0.17.1',
#  'inference_library': 'pymc',
#  'inference_library_version': '5.10.4',
#  'sampling_time': 1.4884159564971924,
#  'tuning_steps': 1000}
print(idata2.posterior.attrs)
# {'created_at': '2024-04-17T08:06:43.592711', 'arviz_version': '0.17.1'}

Would be nice to save those