Fix unsupported timedelta unit in Lifestyle

[tamuri]

The unit M in timedelta is unsupported in newer versions of Pandas. There is one such usage in Lifestyle. This should be replaced by, e.g., 28 days or 4 weeks or something else.

@mnjowe Do you have any preference?

TLOmodel/src/tlo/methods/enhanced_lifestyle.py

Lines 660 to 668 in 283dc65

	# get months since last poll
	now = self.module.sim.date
	months_since_last_poll = round((now - self.date_last_run) / np.timedelta64(1, "M"))
	# loop through linear models dictionary and initialise each property in the population dataframe
	for _property_name, _model in self._models.items():
	if _model['update'] is not None:
	df.loc[df.is_alive, _property_name] = _model['update'].predict(
	df.loc[df.is_alive], rng=self.rng, other=self.module.sim.date,
	months_since_last_poll=months_since_last_poll)

[tamuri]

Fixed in #1616