Test failure in test_two_loggers_in_healthsystem when running on Pandas 2.0

[matt-graham]

One test failure still when running with Pandas 2.0 / Python 3.11

____________ test_two_loggers_in_healthsystem[83563095832589325021] ____________
tests/test_healthsystem.py:841: in test_two_loggers_in_healthsystem
    assert summary_hsi_event['squeeze_factor'].apply(pd.Series) \
E   AssertionError: assert {('Dummy:HSI_Dummy', 2010): 113431310.40270479, ('Dummy:HSI_Dummy', 2011): 113431310.40270479, ('Inpatient_Care:Inpatient_Care', 2010): 0.0, ('Inpatient_Care:Inpatient_Care', 2011): 0.0} == {('Dummy:HSI_Dummy', 2010): 113431310.40270501, ('Dummy:HSI_Dummy', 2011): 113431310.40270501, ('Inpatient_Care:Inpatient_Care', 2010): 0.0, ('Inpatient_Care:Inpatient_Care', 2011): 0.0}
E     Common items:
E     {('Inpatient_Care:Inpatient_Care', 2010): 0.0,
E      ('Inpatient_Care:Inpatient_Care', 2011): 0.0}
E     Differing items:
E     {('Dummy:HSI_Dummy', 2010): 113431310.40270479} != {('Dummy:HSI_Dummy', 2010): 113431310.40270501}
E     {('Dummy:HSI_Dummy', 2011): 113431310.40270479} != {('Dummy:HSI_Dummy', 2011): 113431310.40270501}
E     Full diff:
E       {
E     -  ('Dummy:HSI_Dummy', 2010): 113431310.40270501,
E     ?                                            ^^^
E     +  ('Dummy:HSI_Dummy', 2010): 113431310.40270479,
E     ?                                            ^^^
E     -  ('Dummy:HSI_Dummy', 2011): 113431310.40270501,
E     ?                                            ^^^
E     +  ('Dummy:HSI_Dummy', 2011): 113431310.40270479,
E     ?                                            ^^^
E        ('Inpatient_Care:Inpatient_Care', 2010): 0.0,
E        ('Inpatient_Care:Inpatient_Care', 2011): 0.0,
E       }

Originally posted by @matt-graham in #1114 (comment)

[matt-graham]

Looking at the actual numerical differences this could just be due to floating point errors arising from computation of summations in different orders. The relative difference (113431310.40270501 - 113431310.40270479) / (113431310.40270479) is of order 1e-15 which is close to machine epsilon for double precision floats. We may just need to slightly relax the assertion to use something like numpy.allclose.