Standardise on None, np.nan or pd.NA for missing data

[php1ic]

We don't keep track of whether missing values, purposely or otherwise, are saved as None, np.nan or pd.NA.

The pandas docs have an article on this very topic so let's read it and decide what we want to do.

While we are at it, should the halflife values be timedeltas or similar, rather than floats?

[php1ic]

We aren't going to use timedeltas for half lives.

• They do not have the necessary range
• A half-life is not really a measurement of time in the conventional sense, but rather the measurement of a value.

For both of these reasons we will stick with using a float.

[php1ic]

Actioned in #31, nullable pandas types are used throughout so any missing values are pd.NA.